czech technical university in prague faculty of …etg estimation of throughput gain fap femto...

CZECH TECHNICAL UNIVERSITY IN PRAGUE

Faculty of Electrical Engineering

HABILITATION THESIS

2012 Zdeněk Bečvář

Czech Technical University in Prague

Faculty of Electrical Engineering

Department of Telecommunication Engineering

Mobility Management for Small Cells in

LTE-A Networks

HABILITATION THESIS

Ing. Zdeněk Bečvář, Ph.D.

Prague, (September 2012)

Telecommunication Engineering

Acknowledgement

I

ACKNOWLEDGEMENT

I would like thank to all my colleagues and co-workers for their cooperation in

research work. Especially, I thank to dr Pavel Mach for support in writing project

applications, and for cooperation in scientific work and publications.

This habilitation thesis has been performed mostly in the framework of the FP7

project FREEDOM (No. ICT-248891), which is funded by the European Commission.

I would like to acknowledge all colleagues from FREEDOM consortium for fruitful and

enriching cooperation.

Abstract

II

ABSTRACT

Fourth generation mobile networks implement so-called small cells to cover gaps

in signal such as inside buildings to improve users' experienced quality of services. The

small cells can be connected to the core network via either conventional operator's

backhaul or a user's internet connection, such as ADSL. The former one are represented

by microcells and picocells while the later one are known as femtocells. If a user is

moving along the area with dense deployment of the small cells, a user equipment can

be forced to perform frequent handovers. This leads to redundant signaling overhead

and to a degradation of quality of service for users due to short interruption in

communication during handover. This thesis tackles problems related to mobility

management in fourth generation mobile networks with small cells. First, two

innovative solutions for elimination of redundant hard handovers in small cells are

described. As the simulation results show, both proposals on hard handover are able to

improve network performance comparing to existing and competitive proposals.

Nevertheless, to overcome the problem of the handover interruption, the fast cell

selection must be implemented. Therefore, an improvement of a fast cell selection is

proposed to overcome the drop in quality of service for the scenario with femtocells

with limited capacity of backhaul. The proposed algorithm for the fast cell selection

eliminates handover interruption and it also improves user's throughput and reduces

signaling overhead comparing to the competitive proposals. Last, a management

procedure for temporary access of visiting UEs to femtocells with closed access is

proposed. Two options of management communication are designed: in-band and out-

of-band. The out-of-band communication technology leads to higher energy

consumption at all involved user equipments. However, it does not introduce additional

overhead on communication channels as does the in-band approach.

Table of Content

III

TABLE OF CONTENT

ACKNOWLEDGEMENT......................................................................................................... I

ABSTRACT ............................................................................................................................ II

LIST OF TABLES...................................................................................................................V

LIST OF FIGURES ................................................................................................................VI

LIST OF ABBREVIATIONS ..............................................................................................VIII

1 INTRODUCTION ..............................................................................................................1

2 RELATED WORKS...........................................................................................................4

2.1 HARD HANDOVER ............................................................................................................4 2.2 FAST CELL SELECTION .....................................................................................................6 2.3 TEMPORARY ACCESS TO CLOSED FAP ..............................................................................9

3 MOTIVATION AND OBJECTIVES................................................................................10

4 SCENARIOS AND EVALUATION METHODOLOGY ..................................................14

5 HARD HANDOVER FOR SMALL CELLS.....................................................................17

5.1 ADAPTIVE TECHNIQUES FOR ELIMINATION OF REDUNDANT HANDOVERS ........................18 5.1.1 PRINCIPLE OF THE PROPOSED ADAPTATION TECHNIQUES .............................................18 5.1.2 PERFORMANCE EVALUATION OF ADAPTIVE TECHNIQUES .............................................22 5.1.3 RESULTS OF SIMULATIONS ..........................................................................................23 5.1.4 COMPARISON OF PERFORMANCE OF ADAPTIVE TECHNIQUES ........................................31 5.2 HANDOVER DECISION BY ESTIMATION OF THROUGHPUT GAIN ........................................32 5.2.1 NOTATION AND ASSUMPTIONS FOR ETG.....................................................................32 5.2.2 PRINCIPLE OF ETG......................................................................................................33 5.2.3 ANALYTICAL EVALUATION OF ETG PERFORMANCE.....................................................38 5.2.4 EVALUATION OF ETG PERFORMANCE BY SIMULATIONS...............................................44 5.2.5 DISCUSSION OF BACKHAUL OVERHEAD DUE TO ETG HANDOVER .................................46 5.3 CONCLUSION..................................................................................................................47

6 FAST CELL SELECTION ...............................................................................................48

6.1 FCS IN OFDMA NETWORKS WITH SMALL CELLS ............................................................48 6.1.1 SYSTEM MODEL FOR FCS PERFORMANCE EVALUATION ...............................................51 6.1.2 SIMULATION RESULTS .................................................................................................55 6.1.3 DISCUSSION OF RESULTS AND SUGGESTIONS FOR MOBILITY SUPPORT ..........................59 6.2 ACTIVE SET MANAGEMENT ............................................................................................60 6.2.1 PROPOSED ALGORITHM FOR ACTIVE SET MANAGEMENT...............................................61 6.2.2 SYSTEM MODEL FOR EVALUATION...............................................................................66 6.2.3 SIMULATION RESULTS .................................................................................................67

Table of Content

IV

6.2.4 CONTROL INFORMATION FOR THE PROPOSED ACTIVE SET MANAGEMENT .....................74 6.3 CONCLUSIONS ................................................................................................................76

7 TEMPORARY ACCESS TO CLOSED FAP ....................................................................78

7.1 CONTROL PROCEDURE ENABLING ACCESS OF V-UES ......................................................79 7.1.1 GENERAL FRAMEWORK ...............................................................................................79 7.1.2 IN-BAND APPROACH ....................................................................................................81 7.1.3 OUT-OF-BAND APPROACH ...........................................................................................83 7.2 MANAGEMENT MESSAGES FOR VISITOR ACCESS .............................................................84 7.3 IMPACT OF THE TEMPORARY V-UE ACCESS ON THE V-UE'S PERFORMANCE ..................86 7.4 CONCLUSIONS ................................................................................................................88

8 CONCLUSIONS AND FUTURE WORK.........................................................................89

SUMMARY OF RESEARCH CONTRIBUTIONS.................................................................91

REFERENCES.......................................................................................................................93

APPENDIX ............................................................................................................................97

List of Tables

V

LIST OF TABLES

Table 1. Selection of MCS according to CINR [45]............................................................................... 16

Table 2. Simulation setting ................................................................................................................... 23

Table 3. Summarization of performance of adaptive techniques ............................................................ 31

Table 4. Notation of parameters used for description of ETG................................................................ 32

Table 5. Parameters for ETG evaluation .............................................................................................. 39

Table 6. Comparison of ETG performance with competitive algorithms ................................................ 42

Table 7. Simulation parameters for evaluation of ETG.......................................................................... 45

Table 8. Simulation results for corporate scenario................................................................................ 46

Table 9. Procedures for FCS support in OFDMA-based networks with small cells ................................ 51

Table 10. Simulation Parameters.......................................................................................................... 53

Table 11. Average throughput per user for ∆HM = 3dB, Tadd = 3dB, and Tdel = 3dB; 8 Mbps backhaul

capacity.................................................................................................................................. 59

Table 12. Average throughput per user for ∆HM = 3dB, Tadd = 3dB, and Tdel = 3dB; 100 Mbps backhaul

capacity.................................................................................................................................. 59

Table 13. Notation of parameters used for description of the proposed algorithm.................................. 61

Table 14. Parameters and models used for evaluation of active set management algorithms.................. 66

Table 15. Structure of V-UE Request message ...................................................................................... 84

Table 16. Structure of V-UE Response message .................................................................................... 85

Table 17. Structure of V-UE Info message ............................................................................................ 85

Table 18. Structure of V-UE Confirm message...................................................................................... 86

List of Figures

VI

LIST OF FIGURES

Figure 1. Problem related to the dense femtocell deployment. .............................................................. 10

Figure 2. Frequency reuse constrain in case of FCS with FAPs. ........................................................... 11

Figure 3. Route of data to UE in case of FCS with FAPs....................................................................... 12

Figure 4. LTE-A TDD frame structure used in the simulations. ............................................................. 15

Figure 5. Principle of adaptive hysteresis margin. ................................................................................ 19

Figure 6. Cell radius over RSSImin according to ITU-R P.1238 path loss model. .................................... 20

Figure 7. Average amount of handovers over ∆HM,max for determination of optimum RSSImin. ................. 24

Figure 8. Average DL throughput over ∆HM,max for determination of optimum RSSImin............................ 24

Figure 9. Impact of different methods for determination of ∆HM on average amount of handovers.......... 24

Figure 10. Impact of different methods for determination of ∆HM on DL throughput. ............................. 24

Figure 11. Impact of conventional and adaptive HM on amount of handovers for different densities of

FAPs (CINRwin=50). ............................................................................................................ 25

Figure 12. Impact of conventional and adaptive HM on DL throughput for different densities of FAPs

(CINRwin=50)....................................................................................................................... 25

Figure 13. Average amount of handovers over the Street Width for CINR based adaptive HM............... 26

Figure 14. Average DL throughput of UEs over the Street Width for CINR based adaptive HM............. 26

Figure 15. Impact of different ∆HM,min values on the amount of performed handovers............................. 27

Figure 16. Impact of different ∆HM,min values on the downlink throughput.............................................. 27

Figure 17. Impact of different EXP values on the amount of performed handovers. ............................... 28

Figure 18. Impact of different EXP values on the downlink throughput. ................................................ 28

Figure 19. Impact of adaptive WS on the amount of initiated handovers................................................ 29

Figure 20. Impact of adaptive WS on average DL throughput. .............................................................. 29

Figure 21. Impact of adaptive HDT on the amount of initiated handovers. ............................................ 31

Figure 22. Impact of adaptive HDT on the DL throughput of UEs......................................................... 31

Figure 23.Gain obtained by handover to a FAP.................................................................................... 34

Figure 24. Deployment for analytical evaluation. ................................................................................. 39

Figure 25. Impact of mThr on amount of performed handover. ............................................................... 40

Figure 26. Impact of mThr on relative throughput of outdoor user. ......................................................... 40

Figure 27. Optimum mThr over traffic offered by outdoor user. .............................................................. 41

Figure 28. Impact of error in estimation of kc on the amount of performed handovers........................... 44

Figure 29. Impact of error in estimation of kc on throughput of users. ................................................... 44

Figure 30. Example of simulation deployment for evaluation of ETG. ................................................... 44

Figure 31: Possible introduction of Fast Cell Selection into LTE-A architecture. .................................. 50

Figure 32. Simulation deployment and model of a house....................................................................... 52

Figure 33. Interval between mobility events for hard handover and FCS............................................... 55

Figure 34. Average interruption experienced by UEs due to mobility. ................................................... 56

Figure 35. Served throughput of indoor UEs for open and hybrid accesses. .......................................... 58

Figure 36. Served throughput of outdoor UEs for open and hybrid accesses.......................................... 58

List of Figures

VII

Figure 37. Served throughput of cell-edge UEs for open and hybrid accesses........................................ 58

Figure 38. Proposed algorithm for active set management.................................................................... 62

Figure 39. Impact of α on active set size. ............................................................................................ 67

Figure 40. Impact of α on frequency of active set updates. .................................................................. 67

Figure 41. Impact of α on users throughput. ....................................................................................... 68

Figure 42. Impact of α on ratio of users whose requirements on capacity are not fulfilled. .................. 68

Figure 43. Average throughput of UEs during simulation over amount of offered traffic by the UEs;

throughput of: (a) only indoor users; (b) only outdoor users; (c) all users............................. 70

Figure 44. Average amount of changes in active set of individual users per a simulation step; changes in

active set of: (a) only indoor users; (b) only outdoor users; (c) all users. .............................. 71

Figure 45. Average amount cells included in active set for: (a) only indoor users; (b) only outdoor users;

(c) all users.......................................................................................................................... 72

Figure 46. Average ratio of time spent in the state when UEs requested capacity is not fully provided for:

(a) only indoor users; (b) only outdoor users; (c) all users.................................................... 73

Figure 47. Reference scenario for management of visiting users. .......................................................... 78

Figure 48. General outline of the procedure for V-UE entering the CSG FAP. ...................................... 80

Figure 49. Flow of control messages for V-UE access using IB approach. ............................................ 82

Figure 50. Flow of control messages for V-UE access using OOB approach. ........................................ 84

Figure 51. SINR experienced by V-UE if temporary access is not enabled (dashed blue line) and if the V-

UE is enabled to access this FAP (solid red line). ................................................................. 87

Figure 52. SINR experienced by V-UE over distance between MBS and FAP if temporary access is not

enabled (dashed blue line) and if the V-UE is enabled to access this FAP (solid red line)...... 88

Figure 53. Notation for determination of tc limits.................................................................................. 97

Figure 54. Deviation of tc,min and tc,max over relative distance of users’ path from the FAP’s position. .... 98

List of Abbreviations

VIII

LIST OF ABBREVIATIONS

4G Fourth Generation of mobile networks

ADSL Asynchronous Digital Subscriber Line

AS Active Set

CDMA Code Division Multiple Access

CINR Carrier to Interference plus Noise Ratio

CLC Closed subscriber group List Control

CSG Closed Subscriber Group

DL Downlink

ETG Estimation of Throughput Gain

FAP Femto Access Point

FCS Fast Cell Selection

HDT Handover Delay Timer

HM Hysteresis Margin

IB In-band

IINR Interference to other Interferences plus Noise Ratio

IMSI International Mobile Subscriber Identity

LTE(-A) Long Term Evolution (-Advanced)

MBS Macrocell Base Station

MCS Modulation and Coding Scheme

MIMO Multiple-Input Multiple-Output

OFDMA Orthogonal Frequency Division Multiple Access

OOB Out-of-band

PCI Physical Cell Identification

PRWMM Probabilistic Random Walk/Waypoint Mobility Model

QoS Quality of Service

RB Resource Block

RE Resource Element

RSSI Received Signal Strength Indication

S-GW Serving Gateway

SNR Signal to Noise Ratio

SINR Signal to Interference plus Noise Ratio

TDD Time Division Duplex

TTI Transmission Time Interval

TTT Time-To-Trigger

UE User Equipment

UL Uplink

USIM Universal Subscriber Identity Modul

V-UE Visiting UE

WLAN Wireless Local Area Network

WS Window Size

Introduction

1

1 INTRODUCTION

The fourth generation (4G) mobile networks are assumed to be deployed at

frequencies in order of GHz (e.g., 2 or 2.6 GHz). Transmission at such frequencies leads

to higher attenuation of signal propagated from a transceiver to a receiver comparing to

former bands at roughly 0.9 GHz utilized for GSM. To cover potential gaps in coverage

due to heavy attenuation of a signal at higher frequencies, small cells can be deployed.

In general, two types of small cells are distinguished: femtocells and pico/microcells. In

both cases, radius of cells is low, i.e., in order of tens of meters.

The femtocell, denoted as Femto Access Point (FAP), is assumed to be placed in

user's premises (houses, flats) or enterprises. The FAPs are owned by users and also

controlled by users. Their connection to a core network is enabled via a backhaul of

limited capacity and variable quality. Typically, Asynchronous Digital Subscriber Line

(ADSL) is used as the backhaul connection. Generally, three types of user’s accesses

can be provided by the FAPs: open, closed, and hybrid [1]. In the case of the open

access, all users in the coverage of a FAP can connect to it. A benefit of the open access

consists in an opportunity to offload a Macrocell Base Station (MBS) by serving some

users in areas with heavy traffic load or users far from the MBS [2]. On the contrary, the

FAP with closed access admits only users included in so called Closed Subscriber

Group (CSG) list. The CSG list contains identification of all user equipments (UEs) that

can access the FAP. Users not listed in CSG are not allowed to attach to the closed FAP.

Interference in the case of the closed access should be carefully managed in areas with

dense deployment of the FAPs in order to avoid an impairment of the system

performance. A combination of both open and closed accesses is known as hybrid

access. If the hybrid access is considered, a part of capacity is dedicated for the CSG

users and the rest of the bandwidth can be shared by other users. As presented in [3], the

open access provides higher throughput experienced by users when compared to the

Introduction

2

closed one. This fact is emphasized especially for low density of the macrocell users

[4].

The pico/microcells can be also deployed in users’ premises; however, these cells

are supposed rather for deployment in enterprises or public areas [5]. Contrary to

femtocells, the pico/microcells are under full control of the operator. Moreover, the

pico/microcells should be interconnected with operator's backhaul by a high quality link

with sufficient capacity to serve all traffic transmitted over the air.

Dense deployment of small cells introduces new challenges related especially to

interference mitigation for the closed access and users' mobility management for the

open or hybrid accesses [6]. This habilitation thesis is focused on mobility management.

A mobile user is forced to perform handover from a serving cell to a target cell to keep

the quality of service (QoS). If the user is moving close to the area with dense

deployment of small cells, large number of handovers can be performed within a short

time interval. Then, a drop in QoS is introduced due to the short interruption as a

consequence of hard handover. This is notable especially for real-time services. The

amount of handovers can be adjusted by techniques used for elimination of redundant

handovers, such as a hysteresis or a time-to-trigger [7], [8]. Unfortunately, those

techniques considerably decrease user's throughput in networks with small cells [9].

Moreover, an interruption is still observed if a conventional hard handover is performed

as the user is disconnected from a serving cell before a new connection to a target cell is

established [10]. Fast Cell Selection (FCS) can be exploited instead of the hard

handover to suppress the problem of the handover interruption and QoS decrease in the

networks with dense deployment of the small cells. However, an implementation of

FCS to real networks is more demanding and more complex comparing to hard

handover.

This thesis provides two solutions for hard handover that targets on reduction of

amount of handovers to minimize negative impact of handover interruption. At the same

time, both approaches keeping the same or even improved throughput of the users in

the networks with small cells. Furthermore, FCS is evaluated for both femto and

pico/micro cells to show its efficiency in heterogeneous networks with small cells. Also

an algorithm for management of an active set considering amount of consumed radio

resources is proposed to overcome inefficiency of FCS in networks with small cells.

Last, we propose management procedure for admission of a visiting UE to the CSG

Introduction

3

cells. In this thesis, we focus mostly on femtocells as those are more challenging due to

lower quality of the backhaul than pico/microcells. However, all the proposed solutions

for hard handover and FCS are applicable to pico/micro cells as well.

The rest of the thesis is organized as follows. The next chapter describes and

analyzes related works in the area of the mobility management in 4G wireless networks.

Chapter III defines motivation and objectives of this thesis. Methodology and scenarios

used for the performance evaluations are addressed in Chapter IV. Then, Chapter V

provides description and assessment of two proposals on the management of hard

handover. Chapter VI is focused on advanced mobility support by means of FCS.

Chapter VII defines the management procedure for support of a temporary access of so-

called visiting users to the CSG FAPs. The last chapter summarize major conclusions

and defines potential directions for the future research.

Related Works

4

2 RELATED WORKS

This chapter gives an overview of the state of the art of the work related to the

mobility management in the mobile wireless networks. The research contributions

presented later in this habilitation thesis with respect to the presented related works is

also presented in this chapter.

2.1 HARD HANDOVER

A conventional hard handover is based on comparison of signal levels of serving

and target cells. Handover is executed if the signal level of the target cell exceeds the

one of the serving cell. Several techniques such as Hysteresis Margin (HM) [11], [12],

Time-To-Trigger (TTT) or windowing (also known as signal averaging) [11] are

defined to eliminate redundant handovers in conventional networks without small cells.

In the case of using any conventional technique for elimination of redundant handovers

a drop in throughput is introduced. This is due to a short time when the UE

communicates with the serving station even if a potential target station provides channel

of a higher quality. A drop in throughput is even more significant if the conventional

techniques (e.g., HM, TTT, or windowing) are utilized for elimination of redundant

handovers in scenario with the small cells [9]. A modification of the conventional HM

is defined in [13]. The authors evaluate so-called adaptive HM in scenario with

deployed MBSs but without FAPs. The paper assumes exact knowledge of the distance

among an UE and its serving MBS and exact and invariant radius of the MBSs. The

radius of all cells is assumed to be the same. Nevertheless, the radius is slightly varying

in time in the real networks. Moreover, the radius of individual cells is largely different

if the small cells overlapping with macrocells are deployed. Beside, the exact position

of the FAPs is not defined by operators as it is in charge of the user. Thus, the cell

radius of the FAPs cannot be precisely estimated. Therefore technique proposed in [13]

cannot be applied into the networks with small cells and especially with the FAPs.

Related Works

5

The handover mechanism for FAPs considering asymmetry of a transmitting

power of the FAP and the MBS is introduced in [14] and further extended in [15]. This

mechanism compares the level of the average signal received from the potential target

FAP with the absolute threshold value of -72 dB. Besides, the signal of the MBS is

compared with a combination of the signals from the MBS and the FAP. After the

comparison of the individual results, either the MBS or the FAP is selected as the

serving station. This proposal increases the probability of handover to the FAP if this

FAP provides signal above the threshold and if the FAP is deployed far from the MBS.

Otherwise, if the threshold is not met, the handover is performed as in the conventional

way. Unfortunately, the paper provides no solution for the scenario with overlapping

femtocells. As the authors indicate, the proposed algorithm eliminates redundant

handovers if the FAP is close to the MBS. However, overall amount of handovers is

even increased comparing to the conventional approach. The authors also do not

consider limited capacity of the FAP backhaul in evaluations.

The combination of additional parameters, such as user’s speed and QoS

requirements, for improvement of the handover decision is presented in [16]. Although

the number of the unnecessary handovers is reduced, the throughput is also negatively

influenced. Another speed-aware algorithm is proposed in [17]. The authors exploit a

fuzzy-logic system for the handover decision. The similar idea is further elaborated and

extended in [18] where a new fuzzy-logic based handover algorithm with awareness of

the speed is introduced. However, both papers are focused only on the conventional

networks with macrocells while specifics of the small cells are not taken into account.

Another approach eliminating redundant handovers is to adapt the transmission

power of the FAPs. The proposals dealing with power control adjustment to reduce the

number of redundant handovers in femtocells are presented, e.g., in [19], [20], [21]. All

these proposals eliminate majority of the redundant handovers. Nevertheless, the

advantage of the throughput gain due to the utilization of the open or hybrid accesses

(illustrated in [1]) is also distinctively suppressed by the reduction of the FAP’s

transmitting power. Therefore, these solutions are more suitable for the closed access.

The authors of [22] discuss vertical handover between IEEE 802.16e and Wireless

Local Area Network (WLAN) to maximize user's satisfaction. Taking lower cost of the

connection via WLAN into account, the authors suggest keeping the user connected to

WLAN if it provides sufficient capacity to the user. However, the handover decision

Related Works

6

based only on the current bit rate achieved by the UE leads to the redundant handovers

if WLAN's load fluctuates frequently. Moreover, the authors assume invariable

throughput for users no matter what is its relative position with respect to the MBSs and

the WLAN access points. It means a variability of the throughput in dependence on the

distance between the user and its serving and interfering nodes is not considered.

Furthermore, prediction-based algorithms can be exploited for handover to

improve its efficiency (see, e.g., [23], [24], [25]). The prediction-based approaches

reach high efficiency in determination of the target MBS. However, by deployment of

small cells, the prediction accuracy is strongly affected since small cells' radius is very

low and since the small cells overlapping with MBSs. Moreover, even if the prediction

reaches high efficiency in term of high ratio of correctly predicted target cells; the

handover to the estimated target cell can be inefficient if this cell is a small cell. This is

due to a short time spent by the UE under the small cell's coverage or due to limited

capacity of the femtocells backhaul.

The first contribution of this thesis exploits an idea of the adaptive HM and adapts

it to be easily implemented to 4G networks and also to modify the procedure of HM

adaptation to be applicable in 4G networks with femtocells. We propose to utilize

conventionally reported metrics such as RSSI (Received Signal Strength Indicator) or

CINR (Carrier to Interference plus Noise Ratio) for dynamic adaptation of an actual

value of HM. The second contribution related to hard handover is the algorithm for the

handover decision based on a profitability of handover to the FAP. Handover is

performed only if an estimated throughput offered to a UE by the FAP exceeds the

throughput offered by the MBS. Both radio as well as backhaul parameters of the FAPs

and the MBSs are taken into account in the proposed handover decision. Consequently,

the proposed procedure rejects only those handovers to the FAPs that do not introduce

any considerable improvement in users’ throughput. In other words, the purpose of the

proposed handover decision is to reduce amount of initiated handovers to the FAPs with

low profit (or even with loss) for either network (operator) or users.

2.2 FAST CELL SELECTION

Even if all the proposed modifications related to hard handover are somehow able

to improve the network performance, an interruption due to the hard handover cannot be

eliminated. Moreover, a degradation of a channel quality for cell-edge users is observed

Related Works

7

due to heavy interference if the small cells and the macrocells share the same frequency

bands.

To minimize the problem of the handover interruption, FCS can be implemented.

The FCS has been introduced in 3GPP Release 99 as the SSDT (Site Selection Diversity

Transmission) feature (see [26], [27]). In 3GPP Release 99, FCS strongly relies on the

use of CDMA while only the MBSs are considered. Therefore, modifications required

for utilization of FCS in OFDMA networks with small cells should be defined.

In the case of FCS, the AS is defined for each UE. The AS is comprised of several

neighboring cells of the UE. Neighbor cells are added/removed to/from the AS

depending on the signal level measured by the UE [26]. In [28], the authors compare

fractional frequency reuse in a single cell transmission scenario with FCS enhanced by

adaptive Multiple-Input Multiple-Output (MIMO) mode selection in combination with

interference avoidance technique. The investigation is done for the active set

encompassing two and three MBSs. The active set is updated according to the signal

level received from the neighboring MBSs. Consideration of a relation among the signal

levels of neighboring cells is the conventional approach for FCS.

In [29], [30], the authors propose new metric, denoted as IINR (Interference to

other Interferences plus Noise Ratio), for the active set management. In comparison to

the conventional SINR, the IINR does not take the signal level of the serving cell into

account. The measurement of IINR requires no transmission on the Resource Elements

(REs) that are occupied by reference signals of the neighboring MBSs. The IINR

introduces a gain in spectral efficiency for the cell-edge users and simultaneously it

reduces amount of candidate cells reported by the UE. This way, the load in uplink is

reduced while the downlink is unaffected.

The authors of [31] propose a frequency muting for FCS. The muting is applied to

the second strongest cell according to the UE's measurement. As the results show, this

approach can introduce roughly 10% gain in throughput of the cell-edge users

comparing to the single cell transmission. Further gain of additional 10% can be

introduced by a joint processing. However, this is obtained at the cost of much higher

complexity. Further extension of the muting idea is presented in [32]. The authors

propose the adaptive muting based on a capacity calculation and a power allocation

based on a muting mode selection. The muting is applied to all Resource Blocks (RBs)

Related Works

8

to avoid power wasting. Hence, the transmitting power at some RBs is lowered while

the power at some RBs is boosted. Nevertheless, the overall transmission power is kept

as in the conventional case. The muting mode is considered only if the UE’s throughput

is at least double comparing to the non-muting mode. The results show improvement in

throughput by roughly 5.5% comparing to the single cell transmission.

Analyzing an impact on throughput if a new cell is included in an active set is

presented also in [33]. The authors compare the performance in the case when the

candidate cell would be included with the case when it is not. If the gain by the

inclusion of the cell exceeds the predefined threshold, the update of the active set is

performed.

The FCS introduces a gain in throughput especially at the cell edges where the

interference is not marginal as shown, for example, in [33], [34], [35]. All above-

mentioned papers investigate FCS in the scenario with macrocells only. However,

deployment of the small cells introduces several problems related to the limited

backhaul and small cell radius that could negatively influence the performance of FCS

in the networks with small cells. Therefore, we first evaluate performance of FCS and

compare it with the conventional hard handover in the networks with small cells.

Performance is assessed in terms of the management overhead and the handover

interruption.

Moreover, we also propose the algorithm for more efficient management of the

active set respecting specifics of the small cells. Comparing to the listed related work on

active set management, our proposal differs in several aspects. First, we consider

deployment of the FAPs and its related backhaul problems. Large amount of radio

resources of an MBS could be wasted if the MBS would be included in active set

together with a FAP with weak signal. Therefore, comparing to [31], [32], our proposal

is based on evaluation of the impact of the active set enhancement on the amount of

consumed radio resources of the MBS. Further, a limitation of the FAP backhaul

capacity and delay are considered in our proposal. As the FAPs are supposed to be

connected mostly via ADSL connection, the backhaul capacity is significantly lower

than the capacity of the MBS backhaul. Thus, each inclusion of a FAP into the active

set should take the backhaul limitation into account. In addition, FAP backhaul delay is

a new parameter considered when updating active set in our proposal since this delay is

typically higher than the delay of MBS backhaul.

Related Works

9

2.3 TEMPORARY ACCESS TO CLOSED FAP

In LTE-A networks, the list of CSG users is defined by either a FAP subscriber or

an operator and update of this list requires manual modification of the records in CLC

entity (CSG List Control) [36], [10], [37]. In combination with up to four or eight UEs

allowed to be simultaneously included in CSG list per FAP [6], it is not possible to

update the CSG list frequently. This can be a significant limiting aspect in dense

deployment of FAPs due to inflexible management of the CSG list. A frequent update

can be required, for example, if visitors or guests who attend a subscriber of a FAP

would like to access the FAP. If the subscriber is not willing to include this visitor to the

CSG permanently (for example, due to the limited number of CSG members or due to

the limited throughput of the FAP), the subscriber must manually include and remove

the visitor to and from the CSG list. The manual update of a CSG list is inconvenient

and uncomfortable for the most of the users. A solution for enabling more comfortable

access of the Visiting UE (V-UE) to a CSG FAP is presented in [38]. The authors

propose new message flow to handle the management of the CSG list for the V-UEs.

The solution is based on a configuration of records stored in an operator’s CLC server.

Nevertheless, the authors define only a general framework of the procedure with focus

only on the core network management signaling and do not discuss details on initiation

of the access of the V-UE to the CSG FAP and the management procedures at radio

interface.

Our contribution consists in the design of the control procedure for enabling non-

CSG users to temporarily access a CSG FAP. We propose control messages and their

flow at all involved interfaces for access of the V-UE to the CSG FAP. Two various

approaches, in-band and out-of-band, are proposed and discussed.

Motivation and Objectives

10

3 MOTIVATION AND OBJECTIVES

According to originating standards for 4G mobile networks, the small cells are

expected to be deployed in future mobile and wireless networks to improve coverage in

specific areas with low signal quality. By placing additional stations to the network,

new cell boundaries are introduced. Since heavy deployment of the small cells with low

radius is expected in 4G networks, the procedures for the user’s mobility becomes

initiated more frequently (see Figure 1). Therefore, more often scanning of higher

amount of entities in UE's neighborhood must be performed. Moreover, each handover

generates some management overhead and introduces interruption in user’s

communication. All these aspects lead to a drop of user’s throughput and QoS. This is

getting more apparent with dense deployment of small cells. Hence, large and efficient

deployment of the small cells requires optimizing the principles of user’s mobility

support to ensure continuous high level of service quality.

Figure 1. Problem related to the dense femtocell deployment.

Before mentioned weaknesses could be minimized or even fully eliminated by

implementing FCS. However, deployment of FAPs introduces several problems in the


11

active set management that must be solved for efficient selection of the cells to be

included in the active set for FCS in 4G networks with the FAPs. First, in the

conventional FCS with frequency muting, if a UE consumes significant part of the

resources at the FAP (e.g., due to low signal level), the same resources (at the same

frequencies and in the same time intervals) cannot be used by the MBS included in the

same active set. Thus, it could limit the radio capacity of the MBS.

This situation is shown in Figure 2. The active set of the UE1 contains two FAPs

as well as one MBS. If the FAP1 transmits data to the UE1 at frequencies corresponding

to RB #0 to RB #6, those frequencies can be occupied by neither the FAP 1 nor the

MBS. On one hand, the interferences IMBS-UE1 and IFAP2-UE1 are eliminated and less RBs

can be consumed by the FAP1 to serve the UE1. On the other hand, RBs at the

frequencies corresponding to those used by the FAP1 for delivery of data to the UE1 are

wasted. In the case of dense deployment of the FAPs, this can lead to the situation when

the most of the MBS’s resources are disabled from utilization due to its occupation by

the FAPs involved in the active sets of the UEs along with the MBS.

Figure 2. Frequency reuse constrain in case of FCS with FAPs.

Second, if FCS is enabled, user’s data intended for each UE must be routed to all

cells in its active set (see Figure 3). Due to the limitation of FAP backhaul, inclusion of

FAPs in active sets should consider also the backhaul capacity of individual cell

especially if the FAP is inactive in transmission to the UE. This situation is depicted in

Figure 3. Data destined for the UE1 must be routed to both FAPs in the active set of the


12

UE1. Hence, the backhauls of both FAPs are loaded with all data (in our case, with

seven packets). However, only a part of these packets is transmitted. For example, the

packets #2 and #6 are not transmitted by the FAP1 in Figure 3. These packets are

discarded. On the side of FAP2, only two packets out of seven are transmitted to the

UE1. Other five packets are discarded and those only increases load of the FAP2

backhaul. This problem does not occur in scenario with the MBSs only as the MBS

backhaul is of a very high capacity. Nevertheless, the backhaul of FAPs is typically of a

lower quality.

Figure 3. Route of data to UE in case of FCS with FAPs.

Third, the FAP backhaul is also of a variable quality. If two FAPs are in an active

set of a UE, we can assume that those belong to the same operator (otherwise, the FCS

would not be possible as user is usually subscribed only at one operator). If an MBS and

one or several FAPs are included, we have to ensure that data will be ready at the same

time at all FAPs and MBSs included in the active set. In real networks, it means to

increase packet delay to the maximum delivery delay observed among all cells in the

active set as expresses the next formula:

}D,...,D,Dmax{D iiiA,jA,2A,1

= (1)

where iA,jD represents delay of j-th cell included in the active set of i-th UE.

The general objective of this habilitation thesis is to minimize negative impact of

the management procedures for mobility support on the network performance and QoS

experienced by users.


13

Therefore, the first objective is to define algorithms for hard handover decision to

minimize amount of initiated handovers. This way, the QoS of users and overall

networks performance are improved.

The second objective is to investigate possibility of FCS implementation in the

networks with small cells and further, provide enhanced algorithm for the active set

management considering specifics of small cells.

Last goal is to develop mechanism for easy management of CSG list to enable

faster deployment of CSG femtocells. This part is composed of the proposal of new

management messages and their flow for enabling temporary access of visiting users.

Scenarios and Evaluation Methodology

14

4 SCENARIOS AND EVALUATION

METHODOLOGY

The performance evaluation focuses investigation of an impact of the proposed

procedures on the network metric such as network throughput, distribution function of

signal level experienced by users, and amount of initiated mobility events.

All evaluations are done via simulations in MATLAB since it is common and

universal simulation tool used for mobile networks. Moreover, MATLAB enables

simple implementation of wide range of procedures and algorithms. All models for

simulations and for analytical analysis are in line with models conventionally used for

evaluation of 4G mobile networks. These models are summarized by IMT-Advanced

[39].

For simulation of outdoor user's movement, we consider conventionally used

models such as Direct Movement model, also known as multiple moving mobility

model [40]; Probabilistic Random Walk Mobility Model (PRWMM) [41]; Manhattan

Mobility Model [42]. For indoor user's the mobility model is based on [19].

Street layout and deployment of all network entities follow the general

requirements on simulations as defined in [39] and it is aligned also with the latest

recommendations related to the small cells specifics defined by Small Cell Forum [43].

We consider both rural scenario with less density of users as well as corporate scenario

with high density of users [44].

In all investigated areas of UE's mobility in networks with small cells, only the

slow moving users can perform handover to a small cell. Handover of vehicular users is

usually useless, since high-speed users spend only very short time in the small cell due

to its low radius [16].

In all evaluations, we assume the same 20 MHz wide channel is shared by the

MBSs and the small cells. This channel is at 2.0 GHz for LTE-A. Transmission power

of the MBSs and the small cells is set to 46 dBm and 15 dBm respectively.


15

Unified TDD frame structure of LTE-A release 10 is used in all simulations. The

frame is divided into 10 subframes and 20 slots (see Figure 4). The frame duration is

10 ms and uplink–downlink (UL–DL) configuration “1” is chosen. This configuration

splits the frame into four downlink subframes, four uplink subframes, and two special

subframes (SSs). The SS configuration “0” is utilized for all simulations in this

document. Seven symbols (i.e., Normal cyclic prefix) per subcarrier and 12 subcarriers

per one RB are considered. The spacing of subcarriers is ∆f = 15 kHz. The amount of

bits carried in one RE depends on available Modulation and Coding Scheme (MCS),

which is derived according to [45]. Each slot consists of RBs, which are further

composed of REs. The number of RE per RB ( RB

REN ) is defined by the next equation:

symb

RB

SC

RB

RE NNN ×= (2)

where RB

SCN is a number of subcarriers per RB; and Nsymb is a number of symbols

per the subcarrier.

Figure 4. LTE-A TDD frame structure used in the simulations.

Seven symbols per subcarrier and typically 12 subcarriers per one RB are used for

a normal cyclic prefix. The spacing of the subcarriers is ∆f = 15 kHz. The amount of the

bits carried in one RE depends on available Modulation and Coding Scheme (MCS).

The assignment of the MCS is based on the signal quality according to Table 1 (the

values are taken from [45]).

Downlink throughput of a user is furthermore calculated according to the

subsequent formula:

RB

RERBsymb

RB

SCRBDL NnNNnThr ××=×××= ΓΓ (3)


16

where nRB is the number of occupied RBs (depending on a channel bandwidth as

indicated in [46]), and Γ is the transmission efficiency expressed as the amount of bits

carried per symbol.

Table 1. Selection of MCS according to CINR [45]

CINR [dB] MCS

Transmission

efficiency Γ

[bits/symbol]

CINRmin <CINR <= 1.5 1/3 QPSK 0.66

1.5 < CINR <= 3.8 1/2 QPSK 1

3.8 < CINR <= 5.2 2/3 QPSK 1.33

5.2 < CINR <= 5.9 3/4 QPSK 1.5

5.9 < CINR <= 7.0 4/5 QPSK 1.6

7.0 < CINR <= 10.0 1/2 16QAM 2

10.0 < CINR <= 11.4 2/3 16QAM 2.66

11.4 < CINR <= 12.3 3/4 16QAM 3

12.3 < CINR <= 15.6 4/5 16QAM 3.2

15.6 < CINR <= 17.0 2/3 64QAM 4

17.0 < CINR <= 18.0 3/4 64QAM 4.5

18.0 < CINR 4/5 64QAM 4.8

Algorithm specific evaluation metrics and simulation parameters are further

defined in description of individual proposals.

Hard Handover for Small Cells

17

5 HARD HANDOVER FOR SMALL CELLS

Handover can be initiated due to several reasons, for example, to ensure QoS to

users, to improve coverage or to balance load in networks.

To avoid redundant handovers that increase neither network’s nor users’

performance, several techniques modifying condition for the handover decision are

defined by standards or in literature. Mostly used techniques are: HM, windowing (also

denoted signal averaging), and TTT or its enhancement known as Handover Delay

Timer (HDT) [47].

While HM is implemented, the handover decision and initiation is based on a

comparison of one or several signal parameters (e.g., CINR or RSSI) of a serving cell

and a target cell. The handover is initiated if the signal parameter of the target cell

exceeds the signal parameter of the serving cell at least by a hysteresis ( HM∆ ):

HM

Ser

t

Tar

t ss ∆+> (4)

where Tarts and Ser

ts represents the signal quality parameter of the serving and

target cells respectively in the time instant t.

In the case of the windowing, the handover decision is done if the average value

of the observed signal parameter (e.g., CINR, RSSI, etc.) from the serving cell drops

under the average level of the same parameter at the target cell (see formula (5)). The

average value is calculated over a number of samples denoted as Window Size (WS).

WS

s

WS

sWS

1i

Ser

i

WS

1i

Tar

i ∑∑== >

(5)

where Taris and Ser

is represent i-th sample of the level of the observed signal

parameters at the target and serving cells respectively.


18

Implementation of the HDT is based on the insertion of a short delay between the

time when the handover conditions are met and the time when the handover initiation is

executed. This delay is labeled HDT. The handover conditions have to be fulfilled over

the whole duration of HDT to initiate handover. Generally, handover is performed if:

)HDTt,t(tss HOHO

Tar

t

Ser

t +∈< (6)

where HDT represents the duration of the handover delay timer; and tHO is the

time instant when the handover conditions are fulfilled.

These techniques perform well in the common networks without FAPs. However,

their efficiency drops with implementation of the FAPs [9]. To overcome this problem,

two ways of the handover decision improvement are proposed and investigated in the

following subsections: i) adaptive techniques and ii) throughput gain prediction.

5.1 ADAPTIVE TECHNIQUES FOR ELIMINATION OF REDUNDANT

HANDOVERS

First, the proposals on the adaptation of hysteresis, HDT, and WS are described.

Then, all three adaptive techniques are evaluated by means of simulations in MATLAB

and their performance is confronted with the conventional (non-adaptive) approaches.

5.1.1 PRINCIPLE OF THE PROPOSED ADAPTATION TECHNIQUES

In the conventional HM, the level of the hysteresis is constant. The adaptive HM

is based on the modification of an actual HM∆ value according to the position of the user

in the cell. The HM∆ is decreasing with UE moving closer to the cell boarder. It is

presented in the next equation (defined in [13]):

−×= 0;

R

d1max

4

max,HMHM ∆∆ (7)

where max,HM∆ is the maximum value of HM that can be setup (in the middle of

the cell); d is the distance between the serving MBS and the UE; and R is the radius of

the serving MBS cell.


19

The parameters d and R can be easily obtained or determined neither by the

network nor by the UE (see Figure 5). Especially when the FAPs are deployed in the

networks, its exact position is user dependent and it is not known to the operator.

Figure 5. Principle of adaptive hysteresis margin.

Therefore, we propose to replace the parameters d and R by another metric that

can be utilized more easily and efficiently.

The most of the path loss models describe the relation between the distance d of a

UE from a cell and a path loss (PL) in the following way:

)d(logN)f(X~)d(PL 10+ (8)

where X(f) represents the dependence of the path loss model on the frequency and

other terms usually used in the models; and N is the coefficient related to the type of the

environment. Functions X(f) and N are dependent on the individual path loss model.

The level of the received signal strength at a specific distance (RSSI(d)) depends

on the path loss and the transmission power of the MBS (TPst) as defined in the next

formula:

)d(PLTP)d(RSSI st −= (9)

Furthermore, the distance d can be expressed as an exponential function based on

(8) and (9) as follows:


20

( )

)RSSI)f(XTP(N

1

st10

10st

st

10d

)RSSI)f(XTP(N

1)d(log

)d(logN)f(XTPRSSI

−−

=

−−=

+−=

(10)

Considering (10), formula (7) can be modified in the following manner:

−×=

−×=

−×

−−×

−−×

min,HM

EXP

)RSSIRSSI(N

1

max,HM

min,HM

EXP

)RSSI)f(XTP(N

1

)RSSI)f(XTP(N

1

max,HMHM

;101max

;

10

101max

min

minst

st

∆∆

∆∆∆

(11)

where EXP represents the exponent (in the former adaptive HM defined by (7)

equal to 4); and min,HM∆ is the minimum HM that can be set up (in (7) equal to 0). The

parameters EXP and min,HM∆ can influence the performance of the HM adaptation. The

investigation of the optimal setting of both parameters is tackled later in this document.

The cell radius is typically defined as the distance where a minimal allowed level

of RSSI, denoted as RSSImin, is reached. The typical value of RSSI at the cell’s edge

equals to −90 dBm [48]. However, in the case of the FAP, the cell radius is in order of

tens of meters if the ITU-R P.1238 path loss model [49] is considered (see Figure 6).

Note that the wall loss of 10 dB is included at house boundaries in Figure 6. The impact

of the FAPs radius defined by different RSSImin on the redundant handovers elimination

is analyzed later in this chapter.

Figure 6. Cell radius over RSSImin according to ITU-R P.1238 path loss model.


21

In fact, the border of the cells are neither regular circles nor hexagons since the

system is not distance or signal level limited but it is interference limited. Therefore, the

shape of the cells is strongly influenced also by the interference. Hence, we further

investigate impact of implementation of CINR instead of RSSI for calculation of the

actual level of HM∆ . Generally, a signal level influenced by the interference and noise

(IN) can be described according to the next equation:

INRSSIINPLTPCINR st −=−−= (12)

The CINR level is in different range of values than RSSI. Thereupon, it has to be

related to the difference between maximum and minimum CINRs in the observed area.

Thus, the actual HM∆ according to CINR is derived as follows:

−×= −

−

min,HM

EXP

CINRCINR

CINRCINR

max,HMHM ;101max maxmin

minact

∆∆∆ (13)

where CINRact is the actual CINR measured by the UE; CINRmin and CINRmax are

minimum and maximum values in the investigated area respectively.

The actual CINR of a UE can be easily measured during UE’s operation. It is

usually performed with purpose of the handover decision and initiation. However, also

the minimum and maximum CINR values have to be known for the utilization of the

adaptive HM. CINRmin corresponds to the cell radius and to the CINR level at which the

UE is able to receive data. Therefore, it is set up as a fix value for each FAP and MBS.

CINRmax can be determined by two ways: i) measurement of CINR by a FAP at the

point of its location; or ii) monitoring and reporting of CINR by all UEs connected to

the given FAP and than selecting the highest CINR from all known values as the

CINRmax. The first way implies to equip all FAPs with ability to measure CINR. Hence,

it is not furthermore considered in the evaluations. The second approach utilizes the

knowledge of previous CINR values in the area reported by the UEs. Since the channel

is time variant, the time interval for selection of the CINRmax should be determined. The

parameter CINRwin represents a number of the latest samples utilized for CINRmax

derivation. The optimum value of CINRwin is analyzed later in this chapter.

Analogical modification as for adaptive HM can be done for adaptation of WS

and HDT. Even if neither WS nor HDT are directly related to the signal level, both


22

influence the time spent by the UEs under the coverage of individual cells. Therefore,

both influence the time of the handover decision. Due to the UEs movement, the time of

the handover decision is related to the level of the signals received from all neighboring

cells. The derivation of the actual values for both adaptive techniques is defined by the

following equations:

−×= −

−

min

4

CINRCINR

CINRCINR

maxWS;101WSmaxWS maxmin

minact

(14)

−×= −

−

min

4

CINRCINR

CINRCINR

maxHDT;101HDTmaxHDT maxmin

minact

(15)

where WSmax/min and HDTmax/min are maximum/minimum levels of WS and HDT

respectively.

5.1.2 PERFORMANCE EVALUATION OF ADAPTIVE TECHNIQUES

Evaluations of the modified adaptive technique are performed in the deployment

of FAPs and MBSs along a direct street with a width of 8 m and a length of 500 m as

defined in [44]. The vertical and horizontal distances between neighboring FAPs is 20

and 23 m respectively. Two MBSs are deployed 500 m from the middle of the street.

The direct movement mobility model with the speed of 1 m/s is considered for the

determination of the users’ position. During the simulation, the users are equally

distributed over the street width with spacing of 0.2 m. Major simulation parameters are

summarized in Table 2.

Two metrics for the performance evaluation are monitored: i) amount of

performed handovers, and ii) throughput in downlink. The amount of handovers is

obtained as a number of all initiated handovers. It means, if all conditions for the

handover initiation are fulfilled, handover is counted no matter if it is finished or not.

The throughput via wireless interface is supposed to be with no limitation caused

by the FAP backhaul connection since the FAPs are supposed to be connected to the

backhaul through a high speed optical fiber. Full buffer traffic model is assumed in the

simulations to determine maximum throughput of the UEs.


23

Table 2. Simulation setting

Parameter Value

Carrier frequency 2.0 GHz

Channel bandwidth 20 MHz

Noise spectral density -174 dBm /Hz

Transmitting power of MBS/FAP 46 / 15 dBm

Number of MBSs / FAPs 1 / 50

Speed of outdoor UEs 1 m/s

CINRmin −3 dB

Number of simulation drops 25

5.1.3 RESULTS OF SIMULATIONS

Results of the performance of three adaptation techniques are presented in

following subsections. All proposed algorithms are also confronted with the

conventional techniques without adaptation.

5.1.3.1 ADAPTIVE HYSTERESIS MARGIN

Determination of the optimal RSSImin for the evaluation of the actual HM∆ is

shown in Figure 7 and Figure 8. As the best performing RSSImin value should be the one

enabling maximum reduction of the amount of handovers simultaneously with

minimum impact on the throughput. Based on both figures, the derived optimum

RSSImin is equal to −80 dBm. The figures also show that the selection of inappropriate

RSSImin eliminates the positive effect of the adaptive HM on the amount of handovers

(see e.g., the light blue curve with triangle marker for RSSImin = −75 dBm in Figure 7).

Note that the x axis in all following figures in this section represents the actual value of

HM∆ (or WS or HDT) for the conventional HM (or windowing or HDT). In the case of

HM, WS or HDT with adaptation, the x axis expresses max,HM∆ , WSmax or HDTmax (see

equations (7) and (8)).


24

Figure 7. Average amount of

handovers over ∆∆∆∆HM,max for determination of optimum RSSImin.

Figure 8. Average DL throughput over

∆∆∆∆HM,max for determination of optimum RSSImin.

As stated before, the significant weakness of the RSSI based definition of the cell

edge is that the system is largely influenced by the interference. The comparison of

different approaches of actual HM∆ derivation is presented in Figure 9 and Figure 10.

Both figures are analogical to Figure 7 and Figure 8. The optimum interval CINRwin as

well as the comparison with RSSI based method and the conventional fixed (non-

adaptive) HM can be observed from Figure 9 and Figure 10.

Figure 9. Impact of different methods

for determination of ∆∆∆∆HM on average amount of handovers.

Figure 10. Impact of different methods

for determination of ∆∆∆∆HM on DL throughput.

The utilization of CINR can achieve the same efficiency as the determination of

RSSImin while the CINR based approach is not so sensitive to the CINRwin since the

impact of CINRwin on the number of handovers is negligible. Only a very low CINRwin

leads to a decrease in the handover elimination efficiency. From the throughput point of

view, the lower CINRwin is preferred. Nevertheless, its impact is minor. Comparing to

the conventional fixed HM, the proposed solution reaches the same reduction of the


25

number of handovers with lower negative impact on the throughput. According to

previous figures, roughly 25-50 samples can be determined as the optimum length of

CINRwin.

So far, high density of FAPs (50 FAPs along a street with length of 500 m) was

investigated. Figure 13 and Figure 14 show the impact of the adaptive HM on the

throughput and the amount of initiated handovers for lower densities of FAP densities

(40 FAPs and 20 FAPs along a street with length of 500 m). Lower density of the FAPs

increases efficiency of both conventional as well as the proposed algorithms. In terms of

the amount of initiated handovers, the results obtained by both ways are nearly the same

with only marginally higher efficiency of the conventional approach (less than 2% for

low density and high hysteresis). On the other hand, the increase in throughput is

significant even for low density of FAPs and high hysteresis (up to roughly 6%). The

efficiency of the proposed adaptive HM with relation to the conventional one increases

with the density of FAPs. This is important conclusion for the future when a dense

deployment of the FAPs is expected.

Figure 11. Impact of conventional and

adaptive HM on amount of handovers for different densities of FAPs

(CINRwin=50).

Figure 12. Impact of conventional and

adaptive HM on DL throughput for different densities of FAPs

(CINRwin=50).

Figure 13 presents the distribution of an average amount of handovers over the

street width for different levels of min,HM∆ . The number of handovers is average out over

all simulation drops and over the whole street length. The figure contains results for

CINR based adaptive HM for CINRmin = −3 dB and CINRwin = 50. As can be observed,

the amount of handovers significantly rises with the UE getting closer to the middle of

the street since the difference among all CINRs from cells on both sides is very low. On


26

the other hand, the signal received from the FAPs at the same side as the sidewalk along

which the UE is moving is distinctively higher than the signal from other cells.

Therefore, the UE usually performs the handover only among adjacent FAPs. The

elimination of the most handovers at the sidewalks is achieved even if the HM∆ = 2 dB

while the suggested value in the middle of the street is at least 4 dB (but rather 5 or

7 dB).

Figure 14 illustrates the dependence of the average DL throughput over the street

width. The drop in the throughput when the UE is moving closer to the middle is

obvious. The decrease in the throughput results from lower CINR received if the FAPs

on both sides are roughly in the same distance.

Considering the results presented in Figure 13 and Figure 14, the optimum value

on the sidewalks is ∆HM = 2 dB as it eliminates almost all redundant handovers whilst

throughput is not influenced. On the contrary, the optimum value in the middle of the

street should be defined based on the priority either of the elimination of handovers or

of the throughput. As an optimum ∆HM value should be selected roughly 5 or 7 dB. For

this value, the number of the handovers reaches its minimum; however, the throughput

is still decreasing uniformly. The tradeoff between elimination of the redundant

handovers and throughput should be considered in the middle of the street.

Figure 13. Average amount of

handovers over the Street Width for

CINR based adaptive HM.

Figure 14. Average DL throughput of UEs over the Street Width for CINR

based adaptive HM.

As the requirements on the ∆HM,max depends on the position within the street, the

determination of the general optimum value of ∆HM,max in (11) should be done with

respect to the usual distribution of the users along the street. In the most cases, only the

pedestrians are assumed to exploit open/hybrid access since vehicular users spend very


27

short time in the FAP’s cell due to higher speed. Hence, the major part of users should

be placed on the sidewalks. Consequently, the value of 2 – 3 dB for ∆HM,max can be

selected as the optimum value. Nevertheless, the same scenario can express also the

boulevard where users are moving along the whole street width. In this case, ∆HM,max in

range of 5 – 7 dB is more efficient since low values do not eliminate handovers

efficiently enough.

The evaluation of the optimal values of the parameters ∆HM,min and EXP are

performed in the same scenario as all previous simulations. The ratio of the eliminated

handovers and the relative throughput for the determination of the optimal ∆HM,min are

presented in Figure 15 and Figure 16 respectively. The throughput as well as the ratio of

eliminated handovers are constant until ∆HM reaches the ∆HM,min. The selection of ∆HM

over 1dB leads to the significant elimination of handovers; however the throughput is

also decreased at least by 2.5% per 1dB. While ∆HM,min = 1dB, only less than 60% of

handovers are performed (i.e., over 40% of handovers are eliminated) and

simultaneously, absolutely no negative impact on the throughput is noticed. Thus,

∆HM,min = 1dB should be determined as the optimum value since all other values

automatically results into some drop in throughput whereas maximum throughput can

be still achieved for 1dB. Higher efficiency in the elimination of the redundant

handovers while ∆HM,min = 1dB can be reached by selection of proper ∆HM,max.

Figure 15. Impact of different ∆∆∆∆HM,min values on the amount of performed

handovers.

Figure 16. Impact of different ∆∆∆∆HM,min values on the downlink throughput.

Figure 21 and Figure 22 show the results of the amount of performed handovers

and the downlink throughput for the derivation of the optimum EXP value respectively.

As can be observed from Figure 21, efficiency of the elimination of the redundant


28

handovers is influenced only very slightly by varying EXP and it is increasing with

EXP. No improvement is achieved for EXP higher than 4. The efficiency in the

elimination of the redundant handovers is very close to the performance of the

conventional fixed HM for all investigated values of EXP. The impact of EXP on the

UE’s throughput is also only minor especially for EXP ≥ 6. Therefore, the EXP from

range (2, 4) should be determined as the optimum value. Nevertheless, the EXP

influences the performance of adaptive HM only insignificantly and there is a trade-off

between the amount of performed handovers and throughput.

Figure 17. Impact of different EXP

values on the amount of performed handovers.

Figure 18. Impact of different EXP

values on the downlink throughput.

5.1.3.2 ADAPTIVE WINDOW SIZE

As it is depicted in Figure 19, the adaptive WS leads to the significant elimination

of the performed handovers for low number of averaged samples (roughly up to 7

samples). Then the efficiency of the adaptive technique drops down and the handovers

are performed more often. The decreasing efficiency for higher WS is due to the low

radius of the FAPs. Thus, the signal received from the FAP rises and drops rapidly if the

UE is moving. Therefore, the high WS leads to consideration of the samples obtained

long time ago with respect to the small FAP radius and users' speed. These samples

misrepresent the actual WS and thus the handover is initiated in improper places.

The impact of CINRwin is only minor for a short window. The optimum WSmax for

the adaptive WS is roughly 7 samples since the ratio of performed handovers is the

lowest. Further, the efficiency of handover elimination is rising with CINRwin. However,


29

the results for CINRwin equal to 50 and 500 samples are very close to each other at

WS = 7 samples.

The ratio of the eliminated handovers behaves different for the conventional

windowing with fixed amount of the averaged samples. In this case, the amount of the

initiated handovers is continuously decreasing with growing WS. The efficiency

improvement by approximately 6% is achieved if WS is increased from 7 to 25 samples.

Comparing the conventional and the proposed adaptive windowing, Figure 19 does not

proof any benefit in the elimination of handovers by implementation of the adaptive

WS.

Figure 20 presents the impact of WS on the downlink throughput. This figure

shows no considerable difference between the adaptive and the fixed WS size if WS

value is up to 5 samples. Then, the proposed adaptive WS with shorter CINRwin is

preferable since it leads to a gain in throughput.

By combining the results presented in Figure 19 and Figure 20, it can be observed

that the optimum length of CINRwin is roughly 50 samples. Both figures further show

some throughput gain of the adaptive WS. However this gain is at the cost of lower

efficiency in handover elimination. Thus the adaptation of WS is not profitable

comparing to the conventional fixed WS as it only increases complexity of the system

and it does not introduce any considerable improvement in the performance.

Figure 19. Impact of adaptive WS on

the amount of initiated handovers.

Figure 20. Impact of adaptive WS on

average DL throughput.


30

5.1.3.3 ADAPTIVE HANDOVER DELAY TIMER

The impact of HDT adaptation on the amount of handovers and the downlink

throughput is depicted in Figure 21 and Figure 22 respectively. The range of the HDT

values up to 30 s (x axis in Figure 21 and Figure 22) can be considered since only

pedestrians are assumed to perform handover to a FAP. The vehicular users do not

spend enough time in the femtocell to complete whole handover.

Figure 21 shows that the most of handovers is eliminated by HDT equal to 2 s.

Additional prolongation of HDT up to 6 s leads to moderate decrease of the handover

amount. The HDT over 6 s does not eliminate any further noticeable portion of

handovers. CINRwin influences the results only insignificantly if more than 10 samples

are used.

The conventional as well as adaptive HDTs eliminate handovers with the similar

efficiency except the HDT = 2 s. For this value, the conventional HDT outperforms the

adaptive one roughly by 5 %. Nevertheless, the efficiency of the handover elimination

of both adaptive and fixed HDT can be considered as nearly the same for all other

values of HDT.

As can be observed from Figure 22, increasing length of CINRwin decreases users'

throughput. Hence the shorter length of CINRwin is suggested to eliminate throughput

drop. Comparing the fixed and adaptive HDTs, significantly more negative impact on

the throughput is caused by the technique with no adaptation. The adaptive HDT

enables to reach significant gain in the throughput comparing to the conventional one.

The gain noticeably rises with HDT duration.

Considering the results presented in Figure 21 and Figure 22, the optimum

CINRwin is roughly 25 samples and the most efficient length of HDT is between 4 and

6 s. The adaptive as well as fixed HDTs achieve the similar level of the handover

elimination. Nevertheless, the proposed adaptation of HDT enables throughput gain

between 8 % and 13% for the optimum HDT and CINRwin.


31

Figure 21. Impact of adaptive HDT on the amount of initiated handovers.

Figure 22. Impact of adaptive HDT on the DL throughput of UEs.

5.1.4 COMPARISON OF PERFORMANCE OF ADAPTIVE TECHNIQUES

Table 3 summarizes the performance of all three adaptive techniques with relation

to the conventional non-adaptive ones. It is clear that the most profitable is the

adaptation of HDT since it increases the throughput up to 13% while the same

efficiency in the elimination of the redundant handovers as in the case of the

conventional techniques is retained. Also the adaptive HM is profitable; however the

throughput gain is not so significant. In the case of the adaptive HM, the gain in

throughput increases with FAPs density. Contrary to the both previous techniques, the

adaptive WS does not improve network performance since it increases throughput at the

cost of decrease in the elimination of redundant handovers. Therefore, the same results

can be achieved by modification of the parameter WS without adaptation. Optimum

values of EXP belongs to the interval (2, 4). For HMmin, a value of 1 dB is the most

efficient one. As only signal level parameters are considered for the adaptive

techniques, the same procedures can be applied also to pico/micro cells.

Table 3. Summarization of performance of adaptive techniques

Optimal

value

Optimal

CINRwin

Elimination of

redundant HOs

wrt non-adaptive

Throughput gain wrt non-

adaptive

Adaptive

HM 2 – 7 dB 25 – 50 Negligible decrease

0 – 3 % for low FAP density

0 – 6 % for high FAP density

Adaptive

WS ~ 7 samples ~ 50 Decrease Increase

Adaptive

HDT 4 – 6 s ~ 25 Same 8 – 13 %


32

5.2 HANDOVER DECISION BY ESTIMATION OF THROUGHPUT

GAIN

Algorithm using adaptation proposed in the previous section is very simple and

requires neither any significant modification of the current standards nor any advance

computation. In this section, we provide more complex solution for handover decision

that is based on estimation of throughput gain acquired by performing handover to a

FAP. This approach is further denoted as ETG (Estimation of Throughput Gain). The

application of the novel technique involves several assumptions and requirements

summarized in the next subsection.

5.2.1 NOTATION AND ASSUMPTIONS FOR ETG

To easy following the explanation of the ETG procedure, summarization of the

parameters used in the description of ETG is presented in Table 4.

Table 4. Notation of parameters used for description of ETG

Symbol Definition

cc kt ,

Time in Cell. Mean time spent by users in the cell expressed as a time

interval and number of signal level samples respectively. scc tkt ×−= )1( ,

where ts is the channel quality measurement and reporting period.

outHOinHO kk ,, , Index of signal samples respective to the time instant of the handover

decision ( inHOk , ) and of hand-out from the serving FAP ( outHOk , ).

avgfavgb ss ,, , Estimated mean values of the signals received from the MBS and the FAP

in the time interval 2,kkk HO∈ .

FAPC

Maximum capacity of the FAP available for outdoor user’s limited by the

backhaul.

ctUEd , Data prepared for transmission by the UE during tc.

HOg Real gain in the signal level due to performing handover to the FAP.

estHOg , Estimated gain in the signal level due to performing handover to the FAP.

estFAPestBS TT ,, , Estimated transmission rate of the UE if it stays connected to the MBS and

if it performs handover to the FAP respectively.

estHOG , Throughput gain without consideration of CFAP and ct,UEd .

estHOTG , Throughput gain taking CFAP and ct,UEd into account.

Thrγ Relative threshold for ETG handover initiation.

sbps Current bit rate experienced by the UE at the serving station.

Thrm Multiplier of

sbps to determine

Thrγ ;

sThrThrbpsm ×=γ .

min

connn

Minimum amount of connections to a FAP that has to be performed before considering ETG in handover decision.


33

For implementation of our proposal, we assume the FAP's transmitting power set

to a maximum value to maximize profit of the open access. The FAP’s power control

procedure change transmitting power only for purposes of balancing the interference

level in the network. It means, the power control is initiated only in the case of a rapid

change in interference, e.g., due to neighboring FAP’s turn on/off.

5.2.2 PRINCIPLE OF ETG

The principle of the proposed ETG handover can be explained as follows. Let

sb(k) and sf(k) represent the signal levels of the MBS and the FAP respectively. Both

signals are obtained by periodic measurements and reporting of the signals transmitted

by the MBS and the FAP. The signal level received by a UE is influenced by

transmitting power of the MBS (denoted as Pb,Tx) / the FAP (denoted as Pf,Tx), by path

losses (PLb, PLf), and by shadowing, fast fading, or measurement errors expressed by

parameter ub(k) / uf(k) for the MBS / the FAP. Thus, the signal levels can be defined as:

)k(u)k(PLP)k(sbbTx,bb

−−=

)k(u)k(PLP)k(s ffTx,ff −−=

(16)

To eliminate random effects influencing signal level at the UE, the signal

averaging is assumed. Rectangular window 1)k(w = for )ni,i(k w−∈ is considered

for the sake of simplicity. Parameter nw represents the length of the window. The signal

levels used by the UE for the handover decision are obtained according to the next

formulas:

)k(w)k(s)k(sbb

∗=

)k(w)k(s)k(s ff ∗=

(17)

Conventional handover decision is based on comparison of the signal levels

received from a potential target station ( )k(s t ) with the signal level received from the

serving station ( )k(ss ), i.e., handover is performed if:

HMst)k(s)k(s ∆+> (18)


34

where HM

∆ represents hysteresis margin. Signal levels )k(ss and )k(s t

correspond either to )k(sb or to )k(s f depending on a type of handover as follows:

• )k(s)k(s bs = and )k(s)k(s ft = for hand-in (handover from MBS to FAP);

• )k(s)k(s fs = and )k(s)k(s bt = for hand-out (from FAP to MBS);

• )k(s)k(s fs = and )()( ksks ft = for inter-FAP handover (between FAPs).

In the proposed ETG handover procedure, general condition for the handover

initiation is defined as:

ThrHOgg > (19)

where gThr is a predefined threshold for the handover initiation and gHO is a gain in

signal level. The overall profit in the signal level achieved by handover to the FAP

(gHO) is proportional to the area limited by )t(sb and )t(s f from the time instant tHO,in

till tHO,out, as depicted in Figure 23.

Figure 23.Gain obtained by handover to a FAP.

The gain gHO is defined by subsequent equation:

( )∫ −=out,HO

in,HO

t

t

bfHO dt)t(s)t(sg (20)

If discrete signal samples obtained by a periodic measurement are considered, the

user’s gain is expressed as:


35

( )∑=

−=out,HO

in,HO

k

ki

bfHO )i(s)i(sg (21)

where kHO,in and kHO,out correspond to the indexes of the signal samples obtained at

tHO,in and tHO,out respectively.

Parameters )k(sb , )k(s f , kHO,in, and kHO,out must be found to determine gHO.

Parameters kHO,in and kHO,out represent the instants of the UE’s entering and leaving the

FAP respectively. In fact, exact knowledge of kHO,in and kHO,out is not necessary. Only

the difference, in,HOout,HOc kkk −= , is sufficient to be determined. In praxis, the

parameter kc represents a mean time spent by users in the cell of the FAP and it is

expressed by amount of sampling periods.

An inaccuracy of kc determination can be caused by different movement of users

in the cell and by the variable speed of users. Considering low coverage radius of FAPs,

the estimation of the throughput gain should be distinctively more precise comparing to

the MBS since the difference between minimum and maximum time spent in the cell

varies only slightly comparing to MBSs as derived in [50] (see Appendix).

Once kc is derived, an estimation of the MBS’s and the FAP’s signal levels

progress must be done. The estimation means a determination of )k(sb and )k(s f in

interval ( )out,HOin,HO k,kk ∈ . The precise estimation of )k(sb and )k(s f over the whole

interval ( )out,HOin,HO k,kk ∈ is very complicated since both signal levels are influenced

by many random factors. For the sake of simplification and less computational

complexity we propose to estimate the mean signal level received by the UE in the

interval ( )out,HOin,HO k,kk ∈ from the MBS and the FAP. The mean levels of the signals

are denoted as avg,bs and avg,fs . An inaccuracy of the signal level estimation can be

compensated by selection of a proper threshold gThr for performing handover to the FAP

and by its re-adjustment as explained later in this section.

Value of avg,bs is obtained by an extrapolation of )k(sb in the following way:


36

( )∑=

−×−+

×=

+=

in,HO

min

b

b

k

ii max

bb

max

c

s

sHObavg,b

1i

i)i(s)1i(s

i2

k

)k(ss

∆

∆

(22)

where maxi is the number of the samples considered for the extrapolation; and

)1i(ki maxin,HOmin −−= . For the evaluation of gHO,est, it is necessary also to know avg,fs ,

which is calculated in the same way as avg,bs . If both estimated signal levels and kc are

known, the estimated gain gHO,est is derived as:

( )( )avg,bavg,fcTest,HO sstfg −×= (23)

where fT represents a transformation function for selection of appropriate MCS

according to the received signal levels (see, e.g., [45]). The MCS is commonly

determined based on interference. However, the interference is much more variable than

the signal strength. Therefore, we do not consider interference in our proposal and an

estimation of the interference is left for future research that can potentially further

improve the performance of ETG at the cost of an increase in computational

complexity.

So far, a limitation of FAP backhaul capacity was not considered for the

estimation of the gain in signal level (gHO,est). Moreover, the handover should be

performed only if the UE has data to be send during the connection to the FAP. To

incorporate both limiting factors to ETG, gHO,est must be translated to a gain in user’s

throughput (GHO,est) according to the next formula:

( )( )est,BSest,FAPcest,HO TTkG −×= (24)

where TFAP,est and TBS,est are defined in Table 4.

The final estimated throughput gain with respect to the backhaul limitation and

user’s data is expressed by the following equation:

( )est,HOt,UEFAPest,HO Gd,CminTGc

= (25)

Parameters CFAP and dUE,tc are also explained in Table 4. Note that for pico/micro

cells, the CFAP is assumed to be above GHO,est as the backhaul is dimensioned by

operators to be able serve all radio transmissions.


37

Information on the available capacity of the backhaul of the FAP should be

exchanged between the FAPs and the MBS. This information is delayed due to the

transmission via FAP backhaul, which is of a lower quality than the backhaul of the

MBSs. This delay is supposed to be roughly tens of milliseconds, which corresponds to

the typical end-to-end packet delay for ADSL link [51], [52]. Taking into account the

fact that only pedestrians are admitted to the FAPs, the delay of tens milliseconds leads

to only negligible shift in users’ position (tens of centimeters). Hence, the channel

conditions can be considered as stationary during this very short period. Thus, the delay

only postpones the decision on handover for tens of milliseconds and the estimation of

the throughput gain is affected only insignificantly.

For the handover decision, throughput gain must be confronted with a relative

ETG handover threshold ( Thrγ ). The threshold Thrγ is related to the actual bit rate of the

UE (bpss) and it is expressed as the multiple (mThr) of the current bit rate experienced by

the UE at the serving MBS. This can be defined by the following equation:

ThrSThrest,HO mbpsTG ×=> γ (26)

The Thrγ is used for the elimination of handovers to the FAPs, which offer only

moderately higher throughput than current serving station. In this case, handover is not

profitable due to a short break in user's connection and additional signaling overhead

introduced by the handover initiation.

The level of an over/under-estimation of TGHO,est in real networks is

proportionally the same for all FAPs and MBSs as it is calculated in the same way for

all of these entities. Thus the over/under-estimation of TGHO,est can be reduced by re-

adjustment of Thrγ if more/less handovers to the FAPs are desirable, e.g., for the

purpose of an MBS's offloading.

The evaluation of ETG handover conditions can be performed either once when

the conventional handover conditions, expressed in (18), are met for the first time or

continuously during the whole operation of the UE. In our proposal, the evaluation of

ETG conditions is performed continuously. This way, an impact of rapid channel

variations and an inaccuracy in signal levels estimation are reduced since these

phenomena just postpone the handover for a certain time. In order to avoid negative

affection of the accuracy of TGHO,est by the postponing handover due to both factors, a


38

temporary kc,t is used for derivation of TGHO,est. The kc,t is derived from kc by subtraction

of the time interval elapsed since the conventional handover conditions are fulfilled.

In real networks, the determination of kc is done by an observation of the time

spend by all UEs connected to individual serving cells in the past. Therefore, the

minimum amount of finished connections ( minconnn ) to the FAP is defined for each FAP.

This parameter serves as a trigger for utilization of ETG handover. It expresses the

minimum amount of inputs for derivation of kc that must be collected before the kc is

considered as “accurate enough” to be exploited for ETG. Hence, ETG handover is

considered only if the amount of finished connections is equal to or greater than minconnn .

In the case of the UE entering the area where more FAPs meet the conditions for

handover initiation, i.e., more FAPs fulfill (26), the FAP with maximum TGHO,est is

selected as the target one. If no FAP fulfils ETG handover condition defined in (26)

even if minconnn is reached, the MBS is selected as the target station. If the UE enters the

location with more possible target stations before accurate kc for each FAP in the area is

set up (usually at the beginning of simulation or network operation), the selection of the

target station is based on the conventional handover algorithm.

Since the FAPs are partially controlled by their users, an event such as occasional

FAP’s turn-off should be addressed. In this case, the backhaul is used to inform the

MBS and all adjacent FAPs about the change in a neighbor cell list. All adjoining FAPs

should reinitialize the evaluation of kc and disable ETG handover until minconnn is reached.

Nevertheless, this event is assumed to appear very rarely and can be neglected.

5.2.3 ANALYTICAL EVALUATION OF ETG PERFORMANCE

For analytical evaluation, an MBS and a FAP are deployed in the scenario with

mutual distance dMBS-FAP as depicted in Figure 24. The users are moving along a direct

street with random distance from the FAP, denoted as dUE-FAP. The distance dUE-FAP

represents the shortest distance between the UE's movement and the FAP during a

simulation drop. The performance is evaluated for dMBS-FAP varying in range from 100 to

400 m. For each dMBS-FAP, sixty drops with random speed of users, ranging between 0.97

and 1.74 m/s [53], are performed to average out obtained results. The distance dUE-FAP is

equally distributed for each dMBS-FAP.


39

Figure 24. Deployment for analytical evaluation.

The outdoor users generate constant bit rate traffic during the simulations. Besides

that, fixed indoor users are also considered to generate load of 4 Mbps to the FAP. The

hybrid access with fifty percents of overall backhaul capacity assigned to the indoor

users is applied. The rest of the capacity is dedicated to the outdoor users. The full

backhaul capacity is 8 Mbps. In addition, two scenarios (1 Mbps backhaul with no

indoor traffic and 8 Mbps backhaul with no indoor traffic) are evaluated to show the

impact of the backhaul on the performance of the proposal. All major parameters used

for the evaluation are summarized in Table 5.

Table 5. Parameters for ETG evaluation

Parameter Value

Carrier frequency 2 GHz

Resource blocks per channel 100

Channel bandwidth of MBS and FAP 20 MHz

Noise Power Spectral Density -174 dBm / Hz

Wall Penetration Loss 10 dB

Physical layer overhead 25 %

Outdoor UE speed 0.97 – 1.71 m/s [53]

First, an impact of mThr on the amount of performed handovers and on the

throughput of outdoor users are depicted in Figure 25 and Figure 26 respectively. These

figures are presented only to investigate an impact of mThr on the ETG performance.

Therefore, all results in these figures are related to the maximum value obtained for

individual level of offered traffic, and there is no relation to other competitive handover

techniques.


40

The amount of initiated handovers decreases with increase in mThr until a

minimum of the performed handovers is reached. The minimum number of handovers is

equal to the number of handovers that have to be performed since the signal from the

MBS becomes of a very low quality and it would lead to loosing the connection of the

UE to the network. In other words, if no handover would be performed in this situation,

the UE will not be able to transmit data due to high interference from the neighboring

cells. As the results show, the amount of performed handovers depends not only on the

ETG threshold value, but also on the traffic offered by the UEs. For higher traffic load,

a higher multiplier of the current bit rate of the UE, mThr, must be set up to reach

maximum efficiency in the elimination of redundant handover. This is since achievable

gain in throughput is the multiplication of mThr and the current bit rate of the UE, which

is related to the offered traffic.

Contrary, an increase in mThr leads to only minor drop in the user's throughput.

Lowering the throughput is the cost of avoiding the redundant handovers with low gain

for users. This is due to a utilization of the channel, which is not of the best quality

since the UE stays connected to the MBS although the signal from the FAP is better.

Nevertheless, the impact of ETG algorithm on the mean throughput is only marginal (up

to approximately 0.17% for mThr=10 and 4 000 kbps of offered traffic).

Figure 25. Impact of mThr on amount of performed handover.

Figure 26. Impact of mThr on relative throughput of outdoor user.

As the previous results show, the efficiency of ETG depends on the traffic load

offered by the UE and on mThr. Therefore, an optimal performance of ETG is reached by

utilizing appropriate level of mThr with relation to the traffic offered by users. The

optimum threshold value represents the value of mThr at which the most of handovers

are eliminated while the throughput is still affected only marginally. In our case, it is the


41

value when the amount of performed handovers is nearly at its minimum. The optimum

mThr is depicted in Figure 27. For determination of the optimum mThr, the tolerance of

0.5% of performed handovers is considered, i.e., the optimum corresponds to the value

when the amount of handovers does not exceed minimum of the performed handovers

plus 0.5%. As the results show, the higher mThr is profitable for low traffic offered by

the UEs. This is because of the fact that higher mThr with low offered traffic eliminates

all handovers that would lead to only minor gain in throughput. If lower mThr would be

set up, handovers with only minor gain would be also initiated due to low traffic offered

by users. On the other hand, an increase in UE’s traffic decreases optimum mThr since

even low mThr leads to the higher threshold if a user offers more traffic.

The optimum mThr is also influenced by the backhaul capacity and by the indoor

traffic. If more backhaul capacity is available for the outdoor UE, the optimum

performance is achieved for higher value of mThr as low mThr would lead to a lower

efficiency in the elimination of the redundant handovers. For the backhaul of very low

capacity, the high mThr simultaneously with high level of the traffic offered by the

outdoor UE is useless since the backhaul is not able to serve all user's data. Contrary,

the higher amount of the traffic generated by the indoor UE leads to lowering the

optimum mThr. This is due to the consumption of a part of the FAP's radio resources by

the indoor UE. Consequently, fewer resources are available for the outdoor UE and the

gain introduced by handover to this FAP is lower. Therefore, lower value of mThr is

sufficient to eliminate all redundant handovers.

In praxis, the optimum value of mThr can be determined individually for each UE

as backhaul load, indoor traffic, and the UE’s offered traffic are known to the network.

Figure 27. Optimum mThr over traffic offered by outdoor user.


42

Comparison of the ETG performance with the conventional hysteresis and with so

called Moon's algorithm [15] is presented in Table 6. The table shows the ratio of

served outdoor traffic and the ratio of the performed handovers. The ratio of the served

traffic represents a proportion between the traffic load offered by the UE and the real

traffic transferred by this UE. All results are related to the situation when no techniques

for the elimination of the redundant handovers are used (i.e., the conventional handover

algorithm with ∆HM = 0dB). For ETG, the results represent the values corresponding to

the optimum threshold mThr. Note that the impact on the throughput and the amount of

performed handovers is roughly the same for all levels of the offered traffic if the

optimum mThr is set. The values in parentheses show the difference between ETG and

other competitive techniques. Comparing ETG with the conventional hysteresis, the

hysteresis can eliminate significant amount of the redundant handovers; however, it is

associated with noticeable lowering of the user's throughput. ETG is able to eliminate

significant part of the redundant handovers as well. Moreover, the user's throughput is

nearly unaffected since only those handovers that promise marginal profit for the UEs

are eliminated. Therefore, if the same ratio of the redundant handovers is eliminated by

ETG as well as by the hysteresis with ∆HM = 5.25 dB, the gain of more than 2.5% in the

mean user’s throughput is introduced by ETG. Another interpretation is that ETG

eliminates about 13% more handovers comparing to the hysteresis if both techniques

reaches the same throughput (∆HM = 3.1 dB). It means, additional roughly 47% of

handovers are eliminated comparing to the conventional hysteresis.

Table 6 further shows that Moon's algorithm is outperformed by the ETG very

significantly. Moon's algorithm causes significant drop in throughput simultaneously

with lower efficiency in elimination of redundant handovers.

Table 6. Comparison of ETG performance with competitive algorithms

Handover algorithm Served traffic [%] Ratio of handovers [%]

ETG 99.87 65.66

Hysteresis; ∆ HM = 1dB 99.99 (+0.12) 93.16 (+27.50)

Hysteresis; ∆ HM = 3dB 99.97 (+0.10) 79.27 (+13.61)

Hysteresis; ∆ HM = 3.1dB 99.85 (-0.02) 78.72 (+13.06)

Hysteresis; ∆ HM = 3.75dB 99.05 (-0.82) 74.44 (+8.78)

Hysteresis; ∆HM = 5.25dB 97.36 (-2.51) 65.78 (+0.12)

Moon 95.81 (-4.06) 78.72 (+13.06)


43

So far, an exact estimation of the kc based on the perfect knowledge of the cell

radius was assumed. Therefore, an impact of an inaccuracy in the determination of this

parameter has to be evaluated to meat realistic conditions in the real networks. An

inaccuracy is understood as an error in the determination of kc. It can be caused, for

example, by movement of the UEs in different distances from the FAP or by variable

speed of users. Amount of the performed handovers and the UE’s throughput over the

deviation of kc are illustrated in Figure 28 and Figure 29 respectively. The x-axis

represents maximum error in the estimation of kc (denoted as ε) related to the exact

knowledge of the cell radius. The individual error in kc is then defined by uniform

distribution in interval (-ε, +ε). Both figures show that high estimation error lowers the

amount of the performed handovers. This implies that the high ε leads to the

underestimation of the real gain in throughput and thus additional handovers are

eliminated. However, this is at the cost of a drop in user's throughput. Comparing to the

results of the competitive techniques presented in Table 6, the drop in throughput is still

very low. Even if the estimation error is up to ±100%, the relative throughput (or ratio

of served traffic) is decreased by additional roughly 0.85% comparing to the optimum

determination of kc (see Figure 29). The similar results for the drop in the ratio of served

traffic are obtained by the conventional hysteresis with ∆HM = 3.75dB (see the

difference between ETG and hysteresis in parenthesis in Table 6. However, roughly

65% and 75% of handovers are initiated using our proposed algorithm and the

conventional hysteresis respectively. In addition, another 13% of handovers are not

performed due to the error in kc and thus, only 57% is initiated. Therefore, even if error

in kc estimation is in range of ±100%, the ETG eliminates additional 18% of handovers

comparing to the conventional hysteresis if both causes the same drop in throughput.

Both Figure 28 and Figure 29 further show that the impact of the estimation error

on the throughput as well as on the amount of handovers is nearly independent on the

offered traffic loads.


44

Figure 28. Impact of error in estimation

of kc on the amount of performed handovers.

Figure 29. Impact of error in estimation

of kc on throughput of users.

5.2.4 EVALUATION OF ETG PERFORMANCE BY SIMULATIONS

Analytical evaluations show higher performance of the ETG comparing to the

competitive schemes. However, the performance can be influenced by determination of

kc. Additionally, more UEs simultaneously connected to a FAP can influence the results.

Therefore, we perform simulations for multiplied two stripes scenario with 5x5 blocks

of flats (see Figure 30). This multiplication is used to fully exploit UEs mobility in the

observed area. The FAPs density is equal to two FAPs per a block of twenty flats, i.e.,

10% of flats are equipped with a FAP. Flats equipped with FAPs and the FAPs' position

within the flats are generated randomly with uniform distribution.

Figure 30. Example of simulation deployment for evaluation of ETG.


45

Each UE generates constant bit rate traffic of randomly selected level. The level

of the offered traffic for each UE is generated according to lognormal distribution with

mean of 100 kbps over all UEs.

The major simulation parameters are summarized in Table 7.

Table 7. Simulation parameters for evaluation of ETG

Parameter Value


MBS / FAP transmitting power 46 / 15 dB


Number of outdoor UEs 50


Wall penetration loss 10 dB

Noise spectral density −174 dBm / Hz

Speed of outdoor UEs 0.97 – 1.71 m/s [53]

Simulation step 1 s

Simulation real-time 10 800 s

We perform also a simulation of Adaptive HM under the same simulation

scenario and deployment to compare both proposed algorithms. The results observed

from the simulations are summarized in Table 8. The served traffic as well as the ratio

of handovers are related to the situation, when no technique for the elimination of the

redundant handovers is used (i.e., ∆HM = 0 dB and each UE is connected to the best cell

at each time). In other words, 100% of the served traffic or handovers is the value

reached by a simulation run with all techniques for the handover elimination disabled.

Since each UE in the simulation offers different amount of traffic, we set constant mThr

over the whole simulation for all UEs. Levels of mThr equal to 2 and 3 are selected since

those reach similar level of served traffic as the adaptive hysteresis. The constant mThr is

the simplest way of management but it also slightly decreases efficiency of the ETG.

Particular assignment of individual mThr for each UE according to its bit rate should

slightly reduce overall amount of the performed handovers as presented in 5.2.3. Thus

our presented scenario is the worst case scenario form the performance point of view;

however, it is also the simples for implementation.

The results show similar performance of ETG and adaptive hysteresis in term of

the served traffic. Both outperform the conventional hysteresis by roughly 1.5% of the

served throughput. The adaptive hysteresis reaches the same level of the performed


46

handovers as the conventional hysteresis. Therefore, it confirms its profit in throughput

while nearly no change in the amount of the initiated handovers is reached. This

corresponds to the conclusion obtained by the simulations of adaptive hysteresis in

section 5.1 for the direct street scenario. Contrary, ETG can introduce a gain in the

amount of the eliminated handovers even if a gain in the throughput is still ensured. If

the conventional hysteresis with ∆HM = 3dB and ETG with mThr = 2 are compared, ETG

increases the amount of the transferred traffic by roughly 1.4% and additional 24.45%

of handovers is eliminated (i.e., more than double amount of handovers are eliminated).

Comparing the conventional hysteresis with ∆HM = 5dB and ETG with mThr = 3, the

gain in throughput by ETG is nearly 2% and additional 19.64% of handovers is

eliminated (roughly 85% increase in the handover elimination efficiency).

Table 8. Simulation results for corporate scenario

Handover algorithm Served traffic [%] Ratio of handovers [%]

Hysteresis; ∆HM = 3dB 91.93 78.10

Hysteresis; ∆HM = 5dB 85.43 65.36

Adaptive Hysteresis; ∆HM,max = 3dB 93.29 78.12

Adaptive Hysteresis; ∆HM,max = 5dB 86.97 65.66

ETG; mThr = 2 93.28 54.67

ETG; mThr = 3 87.27 45.72

5.2.5 DISCUSSION OF BACKHAUL OVERHEAD DUE TO ETG HANDOVER

The cooperation among the FAPs and the MBSs via backhaul must be established

to use ETG. The cooperation is used for an exchange of information on the FAP

backhaul status to determine maximum available backhaul capacity for the users. Only

this information has to be delivered to the MBSs for ETG purposes and it should be

available at the MBS in the time instant of the handover decision. Therefore, the

reporting of the backhaul status interval should be similar to the reporting period of

channel quality. In LTE-A, the channel quality reporting period can range between 2 ms

and 160 ms [54]. Considering the worst case, the FAP’s load must be reported each 2

ms, i.e., 500 reports per second must be sent to the MBS. The size of the backhaul load

report should be in tens of bites as the report contains only the indoor traffic load and

the maximum backhaul capacity. Therefore, the maximum overall backhaul overhead of

ETG procedure is couple of kbps in the worst case scenario.


47

Further, an overhead can be generated due to the FAPs' switch-off or switch-on.

For this purpose, only a message with FAP's ID is delivered to all neighboring FAPs to

inform them about this event. Even if the amount of neighbors would be high (e.g., tens

of FAPs), still the overhead in kilobits (tens of FAPs multiplied by tens of bits per

message) is generated only very rarely, since frequent turning-on and off the FAP

cannot be expected. Both parts of the backhaul overhead can be neglected considering

the conventional backhaul capacity in megabits.

5.3 CONCLUSION

Two algorithms for elimination of the redundant handovers are proposed. The first

group, adaptive techniques, is based on exploitation of only parameters conventionally

observed and monitored by the network. As the results show, the most profitable is the

adaptive HDT since it increases the throughput up to 13% while the same efficiency in

the elimination of the redundant handovers as in the case of the conventional techniques

is achieved. The adaptive HM also outperforms the conventional hysteresis.

Nevertheless, a profit of the adaptive HM is lower comparing to the gain introduced by

the adaptive HDT. Contrary to the both previous techniques, implementation of the

adaptive WS does not improve network performance. On one hand, the adaptive WS

increases throughput. On the other hand, the gain in throughput is at the cost of lower

efficiency in the elimination of redundant handovers. Therefore, the same results can be

achieved by modification of the parameter WS without adaptation.

The second algorithm is based on the estimation of the UE’s throughput gain

acquired if handover to a FAP is accomplished. This approach is applicable on

handover performed to a small cells due to its low radius. The results show high

efficiency in the elimination of the redundant handovers while only negligible drop in

the users' throughput is observed. As well, the proposed handover decision algorithm

implies nearly no additional signaling overhead transmitted by the FAP to the MBS via

backhaul. Comparing the proposed algorithm with competitive algorithms, the proposed

one provides higher efficiency in reducing the amount of performed handovers while it

enables to keep higher throughput of users.

Fast Cell Selection

48

6 FAST CELL SELECTION

A solution for ensuring the seamless handover consists in soft handover or FCS.

The major difference between both solutions lies in a way of transmission between a

UE and neighboring cells included in its active set. In the case of soft handover, all cells

in the active set transmit data simultaneously and the receiver combines all data in a

macro diversity manner. Contrary, FCS offers means for the UE and/or the networks to

decide, which cell in the UE's active set is really going to send data in the next

Transmission Time Interval (TTI). To that end, FCS selects and updates the best cell for

the transmission at each transmission interval. Thus, the same data are not sent multiple

as in the case of soft handover.

Soft handover is known as a CDMA specific technique, which cannot be ported

into OFDMA-based systems unless particular algorithms are used at the physical layer

in order to achieve cooperation among the MBSs. FCS is actually a technique derived

from CDMA soft handover. Consequently, its implementation into OFDMA-based

system with small cells requires specific modification at physical layer as well. We

focus on FCS since it implicate less complex requirements on UEs than soft handover.

As mentioned before, FCS in networks with small cells introduces new risks

related to the small cell radius and to the limited backhaul (in case of the femtocells).

Therefore, we first evaluate performance of the networks with small cells to show

whether FCS implementation to the networks with the small cells is even feasible and if

a gain in the network performance could be expected.

6.1 FCS IN OFDMA NETWORKS WITH SMALL CELLS

The first requirement that the OFDMA system has to fulfill for FCS is a time

synchronization among cells in the network, as mentioned earlier. Without proper

synchronization, only the conventional hard handover is possible, where the UE needs

to re-synchronize itself on the target cell after each handover. Synchronizing the system

Fast Cell Selection

49

allows to see FCS as a specific case of joint scheduling, where a set of cells collaborate

in such a way that at each TTI, only the best cell in the set can schedule data toward the

UE. In the case of TDD, the time synchronization of the small cells could typically be

derived from the umbrella MBS. Then each cell in the active set needs to receive the

integral data to be scheduled toward the UE. This principle introduces redundancy.

However, it allows reaching high rates of the cell switching, without flooding networks

with handover events.

Because OFDMA systems such as LTE or LTE-A do not address the notion of

soft handover with the active set of cells serving a given UE, a solution is needed to

allow several MBSs or small cells to participate in the active set in such systems.

Once a radio bearer is established for a UE with one cell in the data path, then it

should be possible to add and remove additional contributing cells. This introduces the

notion of a “serving” cell in the active set, which assumes a particular role, as opposed

to a simple contributor cells. The standard handover procedures still apply whenever the

serving cell in the active set is shifted from a source one to a target one. Figure 31

illustrates required modifications for introducing FCS into LTE-A architecture with

small cells.

New procedures should be defined in order to allow including or removing

contributing cells to or from the active set. When a contributor cell is added to the active

set, then the serving gateway (S-GW) should be notified and it should duplicate packets

toward the new cells in the downlink. Similar mechanisms must be proposed for the

uplink so that contributing cells may take over a role in both uplink and downlink. In

this thesis, we focus on downlink.

Fast Cell Selection

50

Figure 31: Possible introduction of Fast Cell Selection into LTE-A architecture.

The serving small cell or MBS should be in charge of adding and removing

additional contributing cells to the active set. Conventionally, this decision is based on

measurement reports received from the UE, as it is the case for the hard handover

decision itself. A novel algorithm for the active set management is proposed later in this

chapter to improve efficiency of FCS.

If the serving cell is shifted from a source one to a target one, then the set of

contributing cells should be delivered from the source cell to the target cell as a part of

the UE context. Once a successful handover has been achieved for a UE, then the target

cell is free to maintain or modify the set of the contributing cells used by the former

serving cell.

A solution should also be proposed in order to let the contributing cells know if

they are elected to schedule data toward the UE for a given TTI. The solutions defined

in the context of 3GPP release 99 are CDMA specific and cannot be applied outside this

context. The most natural solution is to let the serving cell communicate this

information to the contributing cells on the basis of the UE measurement reports. The

serving cell should provide (and update) the list of contributing cells to the UE for this

report. Since we mainly assume slow moving UEs, reporting periodicity may be set low

enough to maintain low overhead. The nature of the signal level measurement reports

should also be a part of the report configuration. In the simplest case, which is also the

Fast Cell Selection

51

most economical one in terms of uplink bandwidth consumption, only the index of the

best cell should be sent. If a maximum number of cells in the active set is, for example,

8 cells, only 3 bits are required for addressing those cells.

The serving cell should exploit FCS measurement reports from the UE in order to

decide, which cell in the active set will actually be in charge of scheduling data to the

UE. Whenever a modification is decided in this respect, the decision should be

communicated to the involved contributing cells. This command from the serving cell to

the contributing cells should just include the ON/OFF boolean value, together with the

reference of the next TTI where this update should be applied.

Table 9 gives a summary of procedures to be added for supporting FCS in current

OFDMA-based systems with small cells.

Table 9. Procedures for FCS support in OFDMA-based networks with small cells

From To Message purpose Message content

Serving cell S-GW Add a contributing cell Identification of cell to be added

Serving cell S-GW Remove a contributing

cell

Identification of cell to be removed

Serving cell UE Define/update FCS

measurement report

Measurement period

Measurement content as an index in

pre-defined list

UE Serving cell FCS measurement report Measured value

Serving cell Contributing

cell

FCS command ON/OFF value

Identification of next bit of data to

be sent (if the contributing cell is

turned on)

In the following subsections, the performance of the networks with small cells is

evaluated to show whether FCS implementation to the networks with small cells is even

feasible and if a gain could be expected.

6.1.1 SYSTEM MODEL FOR FCS PERFORMANCE EVALUATION

From the performance evaluation point of view, a difference between the femto

and pico/microcells, consists in the capacity of the small cell backhaul. To avoid of

mixing all terms, we use the term "small cell" with meaning of the "femtocell" and

"pico/microcell" if the backhaul is limited and unlimited respectively. By unlimited

backhaul is understood the backhaul able to serve all radio traffic; it means, if the

Fast Cell Selection

52

backhaul capacity exceeds the radio capacity. In our simulations, the "unlimited”

backhaul is represented by the backhaul capacity of 100 Mbps.

We assume co-channel deployment of the small cell and MBSs, i.e., all small cells

shares the same frequency bandwidth as the MBSs. This deployment is more

challenging in term of interference mitigation as all cells interfere to each other.

Furthermore, co-channel deployment is more efficient in spectrum usage (higher reuse

of frequencies).

For the evaluation, a rural scenario with fifty randomly deployed houses within an

MBS is considered according to recommendations defined by the Small Cell Forum

[43]. All houses are of a square shape with a size of 10x10 meters as depicted in Figure

32. Each house is equipped with one randomly deployed small cell and one indoor UE.

The indoor UE moves in line with the probabilistic waypoint mobility model based on

[55]. For this model, several points of stay and a point of decision are defined. In the

point of decision, the indoor UE randomly chooses a point of stay with equal probability

for all points. The time spent in the point of stay is generated according to the normal

distribution taken over from [55]. Beside indoor UEs, also one hundred outdoor UEs are

randomly dropped in the simulation area. All outdoor UEs follow PRWMM [41] with a

speed of 1 m/s.

Figure 32. Simulation deployment and model of a house.

The channel models are also based on the recommendations of Small Cell Forum

presented in [43]. The path loss is modeled according to ITU-R P.1238 and Okumura-

Hata for communication with small cells and macrocells respectively. The channel in

simulations is influenced also by shadowing with a standard deviation of 8 dB and 4 dB

for MBSs and small cells respectively. The transmitting power of the MBS is set to 46

Fast Cell Selection

53

dB while the small cells transmit with 15 dB. Wall losses of 10 dB and 5 dB per outer

and inner walls are also considered.

To minimize effects of randomness of all models, ten simulation drops with a

duration of 7200 s of real-time per a drop are evaluated and averaged out.

For data transmission, TDD LTE-A physical layer is implemented (see Section 4).

Each user (outdoor as well as indoor) offers a constant bit rate traffic during the whole

simulation. User's data are served in a manner that the bandwidth is fairly allocated to

provide the same throughput for all users. For the open access, indoor as well as outdoor

users share the radio resources and the backhaul of the small cells with equal priority.

On the other hand, for the hybrid access, a half of the radio and backhaul capacities is

reserved for the indoor UEs. All outdoor UEs then share the rest of the available

capacity. The major transmission, channel, and simulation parameters are summarized

in Table 10.

Table 10. Simulation Parameters

Parameter Value

Frequency band 2 GHz

Channel bandwidth for macro/small cell 20/20 MHz

Transmitting power of macro/small cell 46/15 dBm

Height of macro/small cell/UE 32/1/1.5 m

Std. deviation of shadowing of MBS/FAP 8/4 dB

Loss of outer/inner walls 10/5 dB

Noise density -174 dBm/Hz

LTE-A physical layer overhead 25%


Number of macro/small cells 1/50

Number of indoor/outdoor UEs 50/100

Number of simulation drops 10

Duration of a simulation drop 7200 s

Several metrics are defined for the performance evaluation: frequency of mobility

events, handover interruption ratio, and served throughput for indoor, outdoor, and cell-

edge users.

The frequency of mobility events is expressed as the mean interval between two

hard handovers or two AS updates. For the hard handover, a mobility event is detected

if the handover is performed, i.e., if the next formula is fulfilled:

Fast Cell Selection

54

( ) ( ) HMst tsts ∆+> (27)

where ts and ss represents the signal level measured by the UE from the target

and the serving cells respectively. In the similar way, an event for the FCS is

conditioned by fulfilling one of the following equations [26]:

( ) ( ) ( ) ( )( ) ( ) ( ) ( )

deltsdelts

addtsaddts

T1ts1tsTtsts

T1ts1tsTtsts

>−−−∧≥−

>−−−∧≤− (28)

where Tadd and Tdel represents threshold for adding and removing cells from the

AS respectively. In the simulations, we set Tadd = Tdel as it is the most common setting

in practice.

The handover interruption ratio is understood as the ratio of the time spent by the

UEs in the state of the interruption due to handover to the overall simulation time. This

is expressed by the next formula:

∑=

=HOn

0h

h

sim

r,HO it

1I (29)

where ih stands for the duration of the interruption introduced by the h-th

handover or the h-th AS update; and tsim is the overall time of the observation (i.e., the

simulation time). It is worth to mention that the interruption in the case of FCS occurs

only if one cell is included in the AS of the UE and if the serving cell of the UE is going

to be switched. We assume the interruption with duration meeting an IMT-Advanced

recommendation for 4G networks. Therefore, we set the interruption to 25 ms.

Served throughput represents the amount of really transferred users' data. It is

observed for indoor, outdoor, and cell-edge users. The indoor users are all users located

inside the houses (50 indoor UEs in the simulations) while the outdoor are all other

users (100 UEs in our simulations). The cell-edge UEs are the users positioned close to

the border of two neighboring cells. According to [32], we define the cell-edge UE as

the user with the level of the signal from the second strongest cell (s2) within the

threshold Tcell_edge (in the simulations, equals to 1 dB) from the signal level of the

strongest cell (i.e., the serving cell, ss) as shown in the subsequent equation:

Fast Cell Selection

55

edge_cell2s Tss <− (30)

The amount of the cell-edge users varies in time depending on the users’ location.

Nevertheless, the trajectories of the UEs are the same for the evaluation of hard

handover and FCS. Thus, the amount of the cell-edge UEs is the same for both as well.

6.1.2 SIMULATION RESULTS

This section presents the results of hard handover and FCS obtained by the

simulations performed in MATLAB.

The impact of ∆HM (for hard handover) and Tadd, Tdel (for FCS) on the frequency

of the mobility events is depicted in Figure 33. The frequency of the events is

proportional to the overhead due to the user's mobility (an overhead related to one

handover or to one AS update is in order of kb [10]). As the figure shows, FCS

introduces more events (shorter mean interval between two events) than hard handover.

For hard handover, the amount of the events is notably reduced by higher ∆HM.

Contrary, the thresholds Tadd and Tdel for FCS decrease the number of the events

negligibly. Nevertheless, the overhead due to the UE's mobility is still insignificant

since an update of the AS (i.e., few kilobits) is required less than once per 340 s even

for very low thresholds. Note that neither access mode (open/hybrid) nor capacity of the

small cell backhaul influence the amount of handovers since it depends only on the

relation between the signal levels of the neighboring stations for the conventional

handover and FCS.

Figure 33. Interval between mobility events for hard handover and FCS.

Fast Cell Selection

56

User QoS is influenced also by the interruption due to handover. The ratio of the

time spent by the UEs in the interruption to the overall simulation time is depicted in

Figure 34. The overall interruption time in the case of hard handover is decreasing with

∆HM as less handovers is performed. However, the hard handover interruption is

significantly higher than the one accounting to FCS. FCS is able to fully eliminate the

interruption even for very low thresholds. The interruption is critical for real-time

services (speech or video calls) as QoS perceived by users is degraded heavily. For non-

real-time services, an impact of the interruption is nearly undetectable by the users as it

is presented only by negligible lowering of bit rate for a very short time (up to 25 ms for

4G networks [56]).

Figure 34. Average interruption experienced by UEs due to mobility.

The amount of the served throughput over the level of the traffic offered by

individual types of UEs is depicted in Figure 35 - Figure 37. Each figure consists of two

subplots showing average throughput for open (left plots) and hybrid (right plots)

accesses. All figures contain results for the backhaul with limited capacity of 8 Mbps

(solid lines) and unlimited backhaul with capacity of 100 Mbps (dashed lines).

The figures confirm the fact that an increase in ∆HM for hard handover lowers the

throughput. This is caused by keeping the UEs connected to the serving cell for a longer

time even if a target cell is able to provide a channel with higher quality. For FCS, an

impact of the thresholds depends on the type of the access and the backhaul capacity.

For the unlimited backhaul, throughput increases with Tadd and Tdel for the outdoor UEs

as more cells are included in the AS and interference experienced by the outdoor UEs is

lowered. For the indoor UEs served by the open access cells with the unlimited

backhaul, the throughput is limited by the backhaul and the positive impact due to an

Fast Cell Selection

57

increase in Tadd and Tdel is negligible. If the backhaul is limited, higher Tadd and Tdel

decrease the throughput of the indoor UEs in the case of the open access. It is a cost of

sharing the backhaul with more outdoor UEs who experience slight increase in

throughput. Nevertheless, this rise in throughput of the outdoor UEs is limited by the

backhaul capacity. For the hybrid access with the limited backhaul, an impact incurred

by Tadd and Tdel is negligible due to a fixed allocation of the resources among the indoor

and outdoor UEs.

According to Figure 35, FCS is profitable for the indoor UEs if a sufficient

backhaul capacity (100Mbps) is provided for both the open and hybrid accesses. If the

backhaul is of a limited capacity (8 Mbps), FCS introduces a heavy loss in throughput

of the indoor UEs for the open access. This loss is a result of fair sharing the small cell

backhaul capacity with outdoor UEs. If the small cell provides higher channel quality

than the macrocell, each UE is trying to transmit data via the small cell, but the

backhaul is not able to serve the data. The hybrid access with the limited backhaul

reaches the same performance for both FCS and hard handover as the fixed ratio of the

backhaul capacity is reserved for the indoor UEs.

Performance of the outdoor UEs is influenced in more positive way by FCS (see

Figure 36). Again, FCS is profitable for all levels of the offered traffic and both

accesses if the small cell backhaul is unlimited. For the limited backhaul, FCS increases

throughput for the offered traffic up to 2 and 1.5 Mbps for the open and hybrid accesses

respectively. Again, the gain of hard handover for high level of the traffic and the

limited backhaul is caused by sharing the resources with more UEs in the case of FCS.

Throughput of the most critical set of users, cell-edge UEs, is depicted in Figure

37. The set of the cell-edge users mostly consists of the outdoor UEs; thus, the behavior

of the throughput of the cell-edge UEs follows the results for the outdoor UEs.

Therefore, FCS is profitable if a small cell is connected via the unlimited backhaul. If

the backhaul capacity is limited, FCS outperforms hard handover only for lower offered

traffic like in the case of the outdoor UEs.

Fast Cell Selection

58

Figure 35. Served throughput of indoor UEs for open and hybrid accesses.

Figure 36. Served throughput of outdoor UEs for open and hybrid accesses.

Figure 37. Served throughput of cell-edge UEs for open and hybrid accesses.

Fast Cell Selection

59

Average throughputs observed from the results presented in Figure 35 - Figure 37

are summarized in Table 11 and Table 12 for the limited and unlimited backhauls

respectively. The numbers in parenthesis represent a gain/drop in throughput introduced

by FCS with relation to hard handover. It can be observed that the average throughput is

improved by FCS in the case of the unlimited backhaul. If the backhaul capacity is

limited, hard handover is more efficient in term of throughput.

Table 11. Average throughput per user for ∆∆∆∆HM = 3dB, Tadd = 3dB, and Tdel = 3dB; 8 Mbps backhaul capacity

Served throughput [kbps]

Indoor

UEs

Outdoor

UEs All UEs

Cell-edge

UEs

Hard HO Hybrid 2510.5 190.01 2700.5 1103.2

FCS Hybrid 2500.8

(–0.4%)

154.52

(-18.7%)

2655.3

(–1.7%)

661.02

(–40.1%)

Hard HO Open 3097.6 192.44 3290.0 1110.7

FCS Open 1615.4

(–47.8%)

172.98

(-10.1%)

1788.4

(–45.6%)

836.83

(–24.7%)

Table 12. Average throughput per user for ∆∆∆∆HM = 3dB, Tadd = 3dB, and Tdel = 3dB; 100 Mbps backhaul capacity

Served throughput [kbps]

Indoor

UEs

Outdoor

UEs All UEs

Cell-edge

UEs

Hard HO Hybrid 3144.7 207.28 3351.9 1111.6

FCS Hybrid 3186.6

(+1.3%)

247.81

(+19.6%)

3434.4

(+2.5%)

1557.5

(+40.1%)

Hard HO Open 3144.7 207.28 3351.9 1111.6

FCS Open 3171.6

(+0.9%)

264.16

(+27.4%)

3435.8

(+2.5%)

1677.4

(+50.9%)

6.1.3 DISCUSSION OF RESULTS AND SUGGESTIONS FOR MOBILITY SUPPORT

Several general remarks and suggestions can be derived from the performed

simulations. First, the performance of hard handover and FCS is influenced by

hysteresis and thresholds as follows:

• Hard handover: throughput decreases with rise of ∆HM disregarding the

backhaul of the small cell.

Fast Cell Selection

60

• FCS: throughput increases with the thresholds for the unlimited backhaul,

while it slightly decreases for the limited backhaul.

Following remarks belong to the limitation of the small cells backhaul:

• Unlimited backhaul capacity: FCS always outperforms hard handover.

• Limited backhaul capacity: FCS is profitable only for the outdoor UEs

offering lower traffic level (1.5 and 2 Mbps for the hybrid and open accesses

respectively).

Last, FCS is profitable for delay sensitive real-time services such as voice calls as

it eliminates the problem of handover interruption.

According to the above mentioned, the backhaul influences the performance of

FCS and hard handover. In related works focused on macrocells only, FCS outperforms

hard handover in all cases (see e.g., [33], [32], [29]). This conclusion is confirmed by

our results for the pico/micro cells with unlimited backhaul. However, an efficiency of

FCS can be degraded more than efficiency of hard handover if the backhaul capacity is

limited. Then, FCS is even outperformed by hard handover for high traffic load.

Therefore, we suggest employing FCS only in the case of low traffic offered by the UE.

For heavy traffic offered by the UE, the UE should perform hard handover to a target

cell if this cell is of the limited backhaul and it is not able to serve the UE according to

its requirements. If the target cell is of the unlimited backhaul, this cell is just included

to the active set along with the current serving cell to reduce interference. Inclusion of

this cell should be performed as soon as possible and the cell should be kept in the

active set for a longer time. This can be easily achieved by setting higher Tadd and Tdel

(for example, 5 dB).

6.2 ACTIVE SET MANAGEMENT

As the previous section show, FCS can be efficient even in networks with small

cells; however, the backhaul capacity needs to be considered in active set management.

The proposed solution for a selection of the active set members (i.e., how to determine

when a cell should be added/deleted to/from active set) is based on the calculation of

amount of consumed radio resources and on backhaul quality consideration. As shown

in Section 6.1, pico/micro cells outperforms conventional hard handover even with the

Fast Cell Selection

61

conventional active set management. Therefore, we focuses on femtocells only in this

section. Note that the same approach can be applied also to the micro/pico cells.

6.2.1 PROPOSED ALGORITHM FOR ACTIVE SET MANAGEMENT

The proposed algorithm on the selection of proper members of the active set

compares the current amount of the consumed radio resources of an MBS with the radio

resources of the MBS consumed if a cell would be added/removed to/from the active

set. In addition, the backhaul limitation, in term of the limited capacity and higher

delay, is introduced in the proposed active set management procedure.

To easy following the explanation of the proposed algorithm, summarization of

parameters used in the description of the active set management is in Table 13.

Table 13. Notation of parameters used for description of the proposed algorithm

Symbol Definition

iN List of neighboring cells of i-th UE, },...,,{ 21

i

nc

iiiiNNNN = .

iA List of cells included in the active set of i-th UE, },...,,{ 21

i

ac

iiiiAAAA = .

ii acnc , Number of cells included in the neighbor cell list and in the active set

respectively.

i

Aj

i

Aj ii RR∉∈

, Amount of the radio resources consumed by the i-th UE if cell Cj is included

in i

A and if it is not included in i

A respectively.

α Gain required for inclusion of a cell intoi

A .

sj DD , Delay of data delivered though cell jC , and maximum acceptable delay for

the service experienced by the i-th UE.

regi

avj bb ,

, , Available capacity of the backhaul of jC and the capacity required by the i-th

UE respectively.

i

S

i

AjjTT i ,, ∈

Throughput of the i-th UE if jC would be added to

iA and throughput

experienced by the i-th UE from the current serving cell.

i

jκ Gain in amount of the MBS's radio resources released by inclusion of jC

intoi

A related to the requested capacity.

Let }UE,...,UE,UE{UE u21= denotes a set of u users in the networks and

}C,...C,C,...,C{C fm1mm1 ++= represents the set of fmk += cells in the network,

where m and f is the amount of the MBSs and the FAPs respectively. Further,

}N,...,N,N{N i

nc

i2

i1

ii= represents the set of the neighboring cells of i-th UE. Each iN

Fast Cell Selection

62

consists of inc neighboring cells. The set }A,...,A,A{A i

ac

i2

i1

ii= is composed of cells

included in so-called active set of i-th UE. Note that i

A is always a subset of iN , i.e.,

ii NA ⊆ . The amount of cells included in the active set of UEi is denoted as iac . The

parameter iac is known as active set size.

The principle of the proposed algorithm is depicted in Figure 38. A new cell is

included into iA , if all defined conditions are met. The cell is removed from the

existing iA if at least a condition is not fulfilled.

Figure 38. Proposed algorithm for active set management.

If ij NC ∈ and i

j AC ∉ , then the cell can be included into the iA if:

i

Aj

i

Aj ii RR∉∈

<α (31)

where i

Aj iR∈

represents the amount of the MBS's radio resources consumed by the

iUE if the jC would be included in the iA , i

Aj iR∉

represents the amount of the MBS's

radio resources consumed by the iUE if the jC would not be included in the iA , and α

Fast Cell Selection

63

represents a gain required for the inclusion of the jC in iA . The MBS's resources are

considered in this equation rather than the FAP's resources since each FAP is supposed

to serve only low amount of users comparing to the MBS. Thus, any change in an active

set influences large amount of the macrocell users but only couple femtocell users.

Both i

Aj iR∈

and i

Aj iR∉

are derived from the reports on signal quality (e.g. SNR)

measured by the iUE from all cells included in iN (see, e.g., [32]). If SNR of all cells

included in iN is measured, SINR can be determined. Then, SINR is mapped to a

modulation and coding scheme (MCS) according to, for example, [45]. Each MCS

defines a modulation and a coding rate. Therefore, an amount of bits in a RE, denoted as

REb , can be derived as a multiplication of the coding rate (cr) and amount of bits per

symbol of the modulation (bps), i.e., bpscrbRE ×= . Knowing amount of the radio

resources required by the iUE and REb of appropriate channel between the iUE and the

jC , the amount of the consumed resources is determined as a simple ratio of data

intended to be sent by the UEi ( UEd ) and REb ; REUEi bdR = . Difference in derivation of

both i

Aj iR∈

and i

Aj iR∉

consists in consideration of the jC in the interference evaluation.

For i

Aj iR∈

, the signal from the jC is not taken into account since no cell included in the

iA can transmit at the same frequencies as the serving cell. Contrary, the signal from

the jC is included in the interference for i

Aj iR∉

.

Once the inclusion of the jC in the iA is profitable from the amount of consumed

radio resources of the MBS point of view, the quality of the backhaul of the jC is

evaluated. A problem of a packet delay due to transmission via the backhauls with

different quality is fixed as follows. To cope with the delay, we suggest an additional

condition for inclusion of a cell into the iA as defined by the next formula:

i

sj DD ≤ (32)

where, jD is the delay of data delivered though jC , and isD is the maximum

acceptable delay for the service experienced by the iUE . Note that this problem is

common problem of the handover procedure. Therefore, it should be considered even in

the conventional hard handover. However, the backhaul of the MBS typically provides

Fast Cell Selection

64

high quality of the connection with low delay and this condition is fulfilled

automatically.

The backhaul of the MBSs is planned to be able to serve all the data transmitted

via the radio interface. It means the bottleneck does not appear on the backhaul of the

MBSs. In the FAPs, the situation is exactly the opposite. Since the FAPs are supposed

to be connected to the networks via a backhaul with limited capacity, previous

conditions (31) and (32) are complemented by additional one focused on the backhaul

capacity. The next condition is considered only if the jC is a FAP. The femtocell jC

can be potentially included in iA only if:

0bb req,i

av,j ≥− (33)

where av,jb and req,ib is the available backhaul capacity of the j-th FAP and the

backhaul capacity requested by the iUE respectively. After fulfilling (33), the jC is

included in temporary active set itempA . The i

tempA is composed of all cells that should be

included in iA as those meet (31), (32), and (33). If only one UE is supposed to newly

include the jC into iA , then the temporary active set can be added to the iA , i.e.,

}A{}A{}A{ itemp

ii += . If more UEs would like to include the jC in their iA , then the

backhaul limit is reconsidered. The cell jC should be added to more active sets only if

the FAP will be still able to serve all UEs as expresses the following equation:

0bbitempAj|i,i

req,i

av,j ≥− ∑∈

(34)

If (34) if not fulfilled, only iA of the selected UEs will be updated. A procedure

for the selection of the most appropriate UEs, whose active set will be enhanced by the

jC should be defined. For this purpose, we define new parameter ijκ . This parameter

represents a ratio of the gain caused by the inclusion of the jC to the requested

backhaul capacity. It is defined by the next formula:

req,i

i

Aj

i

Aji

jb

RR ii ∈∉−

=κ (35)

Fast Cell Selection

65

The jC is included only to the iA of the UEs that leads to the highest ijκ . For this

purpose, the ijκ is reordered in descending order as follows:

}},...}{{},{{)( ij

ij

ijconstj

ij maxmaxmaxsort κκκκ −== (36)

Then the jC is sequentially added to the iA for maxb,...,1i = . The maxb is

determined as )bmax( ; )u,...,1(b = for which the following formula is still valid:

0bbb

1i

req,i

av,j ≥−∑=

(37)

A specific situation, when (31) and (32) are fulfilled while (33) is not can occur.

In this case, a FAP can provide higher throughput even if other UEs with this FAP in

the active set could suffer from the inclusion of the FAP into the iA . Nevertheless, the

drop in throughput of the UEs served by the FAP can be insignificant and the

throughput of all of these UEs can be still above the one provided by the MBS.

Therefore, the cell jC is included into iA even if not enough available backhaul is

provided by the jC . The inclusion is conditioned by fulfilling subsequent equation:

i

S

i

Aj,jTT i >

∈ (38)

where i

Aj,j iT∈

is the throughput of iUE if the jC would be added to iA and iST

represents the throughput experienced by the iUE from the current serving cell. To add

the jC to the iA , all UEs currently served by the jC must still be experiencing higher

capacity than the capacity provided by the MBS.

Description above focuses on conditions and algorithm for inclusion of a cell into

an active set. The opposite case, that is, removal of a cell from the active set must be

defined. In our proposal, the jC is deleted from the iA if it results in lesser

consumption of the MBS's radio resources. In other words, the jC is removed if

condition (31) is no longer met. Of course, the cell is removed if its backhaul capacity

or delay changes and either (32) or (33) becomes not fulfilled.

Fast Cell Selection

66

Selection of a serving cell is based on comparison of the signal levels measured

by the UE. If the cell with the strongest signal measured by the iUE can fully served

this particular iUE , then the cell is selected as the serving. However, if even this cell

cannot provide enough capacity, the cell providing maximum throughput is selected.

6.2.2 SYSTEM MODEL FOR EVALUATION

The transmission power and path loss models follows those used in evaluation of

FCS in section 6.1. This section describes especially parameters, in which both

simulations differ.

Since the main objective is to assess when active set of the UEs should be

updated, the outdoor UEs are supposed to be moving in comparison with the previous

evaluations. The individual parameters, and values set for the evaluation of the proposal

are presented in Table 14. The outdoor UEs are moving according to PRWMM (see

[41]) with a constant speed of 1 m/s. All UEs transmit data according to constant bit rate

model with a bit rate in the range from 60 kbps to 4Mbps. Each UE has different

requirements on the delay for its services. The required delay is selected among three

possible classes: high demands (delay of backhaul ≤ 50 ms), medium demands (delay of

backhaul ≤ 75 ms), low demands (delay of backhaul > 75 ms). The selection of the

delay requirements is done randomly. The probability that UE’s demands on the delay

is high/medium/low is 5%/20%/75%.

Table 14. Parameters and models used for evaluation of active set management algorithms

Parameters Value


MBS / FAP transmitting power 46 / 15 dB


Number of indoor/outdoor UEs 50 / 100


Wall penetration loss 10 dB

Noise spectral density −174 dBm / Hz

Size of simulation area 2000 x 1000m

The FAPs' backhaul is limited to 8 Mbps for downlink and 50 % of its capacity is

supposed to be consumed due to ADSL aggregation and signaling overhead.

Consequently, the real available capacity of the FAP backhaul dedicated for the UEs in

Fast Cell Selection

67

the simulation is 4 Mbps. The delay of each backhaul is selected according to the

measurement in a real network provided by Telkom Indonesia in [52]. To eliminate an

effect of random variables, the simulation duration is set to 3600 seconds of real time

and we run five simulation drops.

6.2.3 SIMULATION RESULTS

The results obtained by the simulations in MATLAB are split into two sub-

sections. The first one presents the results for determination of appropriate α for the

proposed algorithm. The second part of the results shows the comparison of the

proposed algorithm with selected competitive proposals.

6.2.3.1 EVALUATION OF THE PROPOSED ALGORITHM

As Figure 39 shows, the average size of the active set per a UE over the whole

simulation decreases with the amount of the traffic offered by the UEs. This is due to

the limited capacity of the backhaul of the FAPs. Once the backhaul is fully utilized, the

FAP is included into other active set(s) only if it improves the throughput of the UE and

ensures enough capacity even for all UEs with the FAP in current active set. On the

other hand, higher value of α lowers the size of the active set since lower profit must

be achieved at the side of the MBS to include a FAP into the active set (see (32)).

Analogically, the frequency of an active set updates (Figure 40) rises with lowering α

or offered traffic. The active set update rate represents the average amount of changes in

the active set of a user per a simulation step (a second). Each inclusion or removing of a

cell into an active set represents one change.

Figure 39. Impact of α on active set

size.

Figure 40. Impact of α on frequency of

active set updates.

Fast Cell Selection

68

Besides the active set, α influences also user's throughput (Figure 41) and so

called capacity outage (Figure 42). The throughput is affected by α only for higher

offered traffic loads. All cells, even FAPs, are able to serve high amount of users

without reaching a backhaul limit for low offered traffic. Thus, no impact of α on the

throughput is observed. However, if the offered traffic increases, high α leads to the

selection of only considerably profitable cells as candidates to be included in the active

sets. If α is low, each FAP is included in large number of active sets and all users

connected to this FAP must share the limited backhaul. Note that we assume

proportional fair sharing of the FAP backhaul capacity among all users connected to it.

The capacity outage is understood as a time for which a UE’s requirements in

term of throughput are not fully served. In other words, the real transferred capacity is

lower than the traffic offered by the UE during this time. For a low offered traffic, the

capacity outage rises with α . Contrary, the performance is slightly improved for a

higher α or if a heavy traffic is generated by the UEs. This opposite behavior is a result

of a load balancing among individual cells. For a low offered traffic, a higher α limits

exploitation of the available backhaul of the FAP even if the MBS is not able to fulfill

all UE's requirements. On the other hand, for a heavy traffic, low α leads to more FAPs

included in the active sets. Thus, more UEs share the FAP backhaul capacity and the

backhaul limit leads to a higher number of unsatisfied UEs.

Figure 41. Impact of α on users

throughput.

Figure 42. Impact of α on ratio of

users whose requirements on capacity are not fulfilled.

Based on the presented results, α > 2 is considered as the appropriate gain since

maximum throughput and minimum amount of active set updates are generated. For the

Fast Cell Selection

69

purpose of evaluation of the proposal and competitive schemes, α = 2 and α = 3 are

selected.

6.2.3.2 COMPARISON OF COMPETITIVE ALGORITHMS

The proposal is compared with two algorithms: conventional FCS active set

management [27] and with the proposal on capacity based FCS active set management

proposed in [32]. The capacity based FCS is selected for the comparison since this

proposal outperforms any other similar proposals as presented in [32].

As shown in Figure 43, the proposed active set management algorithm improves

throughput of all UEs (indoor as well as outdoor) comparing to the conventional and the

capacity based FCS (Figure 43c). The gain in throughput rises with the amount of

offered data by the UEs and it is nearly independent on the level of α . The throughput

gain for the indoor users (Figure 43a) is up to roughly 28%, 17%, and 41% comparing

to the capacity based FCS, the conventional FCS with ∆HM = 3dB and the conventional

FCS with ∆HM = 5 dB respectively. For outdoor users, a minor gain (up to 4%)

comparing to the capacity based FCS is notable up to 2000 kbps (Figure 43b). Then

both schemes perform similarly. Comparing the proposal with the conventional FCS,

the gain rises up to 20% with the offered traffic up to 2000 kbps. For the offered traffic

over 2000 kbps, the gain gets stable and equals approximately to 7%. The rapid drop

experienced by the proposal and the capacity based FCS at 2000 kbps is due to the FAP

backhaul limitation and it can be explained as follows. For each FAP, a UE is deployed

indoor in our simulation deployment. Therefore, this UE is attached to this FAP most of

the time. Including the FAP in the active sets of other outdoor UE, the backhaul must be

shared by all UEs with this FAP in the active set. Since the available backhaul capacity

is 4000 kbps in average in the simulations, an inclusion of the FAP to an active set of

any outdoor UE automatically limits the transmission capacity up to 2000 kbps.

Fast Cell Selection

70

(a) (b)

(c)

Figure 43. Average throughput of UEs during simulation over amount of offered traffic by the UEs; throughput of: (a) only indoor users; (b) only outdoor users;

(c) all users.

Frequency of active set updates is presented in Figure 44. Each event in the active

set (either inclusion/deletion of a cell to/from active set) is linearly interconnected with

certain amount of a management overhead. Therefore, these figures represent also the

related amount of control overhead generated due to the active set management. In the

case of the indoor users only (Figure 44a), the lowest rate of the active set update is

reached by the conventional FCS. However, it is at the cost of significantly decreased

throughput as presented in Figure 43a. The capacity based FCS and the proposed

scheme performs similarly in term of the active set update rate. The sudden rise in the

case of the proposal is again due to the backhaul limit as explained above for the

throughput. Regarding outdoor users presented in Figure 44b, the results are exactly

opposite. The lowest rate of the active set update is reached by the proposal. The rate of

updates decreases with higher values of α and with an increase in offered throughput

for our proposal. Lower frequency of the active set updates for a higher α or for a

Fast Cell Selection

71

higher offered traffic is due to a lower probability of fulfilling condition (32) or a higher

requirements of all connected UEs on the FAP backhaul respectively. Note that

frequency of the active set updates of the indoor users is roughly ten times lower

comparing to the outdoor users. Thus, the proposed scheme outperforms the

conventional FCS and the capacity based FCS roughly by 58% - 65% and by

20% - 43% (depending on the amount of the offered traffic) respectively if overall rate

of the active set update (indoor as well as outdoor UEs) is evaluated (Figure 44c). Based

on the results, we can stated, that the proposal generates significantly less overhead

comparing to the both competitive scheme.

(a) (b)

(c)

Figure 44. Average amount of changes in active set of individual users per a simulation step; changes in active set of: (a) only indoor users; (b) only outdoor

users; (c) all users.

In Figure 45, the average size of the active set per UE over the whole simulation

is depicted. For the indoor UEs (Figure 45a), only roughly one cell is included in the

active set for the conventional FCS. This cell is typically a local FAP deployed in the

same house as the UE. Signals of other cells’ (either FAPs or MBSs) are usually

Fast Cell Selection

72

attenuated significantly due to intervening walls. Hence, these cells do not provide

signal with sufficient quality to be included in the active set. For the capacity based and

proposed FCS, roughly two cells are included in the active set on average. This is

typically an MBS and the local FAP. Other FAPs provide weak signal (at least two

walls are in between the FAP and the UE) to be included in the active set.

For the outdoor users (Figure 45b), the active set consists of an MBS and several

closest FAPs. The exact number of the FAPs included in the active set depends on the

offered traffic level for the capacity based FCS. In the case of the proposed FCS and the

conventional FCS, the number of FAPs in the active set is further influenced by α and

by hysteresis respectively. The average size of the active set is presented in Figure 45c.

Note that based on standalone size of the active set can be concluded neither lower nor

higher active set size is profitable. This parameter just show typical amount of cells

involved in the active set communication, which can be further used, for example, in

optimization of cooperative communication.

(a) (b)

(c)

Figure 45. Average amount cells included in active set for: (a) only indoor users; (b) only outdoor users; (c) all users.

Fast Cell Selection

73

Last set of figures (Figure 46a - c) presents the impact of individual FCS

algorithms on the ratio of the time, when the user's requirements on capacity are not

fulfilled. Comparing the indoor (Figure 46a) and outdoor (Figure 46b) UEs, the

satisfaction of the outdoor UEs is lower comparing to the indoor UEs. The reason is that

the indoor UEs are usually not limited by the radio capacity. The bottleneck is typically

located on the FAP backhaul, which is of a higher capacity comparing to the radio

capacity of an MBS.

(a) (b)

(c)

Figure 46. Average ratio of time spent in the state when UEs requested capacity is

not fully provided for: (a) only indoor users; (b) only outdoor users; (c) all users.

The profit achieved by the proposal rises with offered traffic load and it is almost

independent on the value of α for the indoor users. The improvement is up to 13%, 7%,

and 12% when compared to the conventional FCS with 3dB hysteresis, the conventional

FCS with 5dB hysteresis, and the capacity based FCS respectively. For the outdoor

UEs, the maximum profit (50%) is reached for 100 kbps of the offered traffic if

comparing the proposal with the conventional FCS. The performance of the capacity

based and the proposed FCS is roughly the same for the outdoor users. Nevertheless, the

Fast Cell Selection

74

proposed algorithm is outperformed by none of both competitive algorithms for all UEs

(indoor and outdoor) and for all offered traffic levels. The efficiency of the proposal

comparing to both competitive schemes consists in more efficient selection of the cells

included in the active sets of individual UEs.

Besides the capacity constrain for the backhaul of the FAPs, also a delay outage

should be tackled. By the delay outage is understood the situation when a UE's

requirements on the delay are not met. Our proposal suppresses the outage delay to the

minimum achievable level. This minimum is given by occurrence of the situations when

none of the neighboring cell is able to provide sufficient delay. However, this problem

is not related to FCS active set management. The delay outage of the competitive FCS

schemes depends on the quality of the backhaul of the FAPs. During our simulations,

the delay outage for both competitive FCS schemes was roughly in range of 1 - 2.5 %

above the outage of our proposed scheme, which reaches the delay outage under 1%

even for heavy traffic load. Note that the delay outage introduced by our scheme is only

due to the overloading of the system by high amount of offered traffic by the UEs.

6.2.4 CONTROL INFORMATION FOR THE PROPOSED ACTIVE SET

MANAGEMENT

To enable FCS in networks with FAPs, exchange of control information among

the FAPs and the MBSs must be defined. Information on the backhaul quality and the

FAP's radio quality must be reported to the serving cell. However, the quality of the

radio channel is periodically reported for common handover purposes. Therefore, no

additional overhead is introduced by the proposed algorithm in term of information on

the radio channel quality.

Each FAP is aware of its approximate backhaul quality as it needs this

information to schedule users' data over the backhaul. Moreover, estimation based on

the latest experienced backhaul quality can be considered. Nevertheless, the information

on the backhaul quality must be delivered to the serving cell, which is supposed to take

control over the handover decision. Thus, each potential candidate for inclusion in an

active set should report the available capacity and the packet delay to the serving cell.

The reporting of the backhaul quality can be included in the control information for the

coordination of the MBSs and the FAPs.

Fast Cell Selection

75

The information on the backhaul delay can be provided in form of the range of the

delays related to the experienced service class. It means, the delay does not need to be

reported as an exact number but only as an index representing appropriate range of the

delays. Therefore, its size is only of several bits. For example, 4 bits enable to

distinguish 16 classes, which is sufficient number, higher than amount of classes used

by IP protocol or in LTE-A. The information on the capacity should be expressed as an

absolute amount of available resources. The number of bits required for this information

depends on accuracy of reporting information. Sufficient amount is 16 bits as it enables

to distinguish 216

levels of the available capacity (for example, it is the resolution of 4

kbps for 16 Mbps backhaul).

Transmission of the information on the backhaul quality can be either triggered by

a handover request or periodical. The drawback of the handover triggered reporting is

an additional delaying of handover (in tens of ms) due to delivering information on the

backhaul status to the serving cell. However, its overhead is negligible since only few

additional bits are transmitted per a handover. On the other hand, the periodic reporting

does not delay handover but it increases signaling overhead. The maximal amount of

the overhead generated during the periodical reporting can be determined as follows.

The bit rate necessary for the reporting can be expressed as:

rep

ri

repT

SBR = (39)

where Sri is the size of the reported information and Trep is the interval between

two reports. The maximum size of a report is 16+4 bits as stated earlier. The minimum

reporting period is supposed to be equal to the frame duration, which is 10 ms in LTE-

A. Then the maximum reporting overhead is 2 000 bps. This amount of the overhead is

still negligible comparing to the expected backhaul capacity in Mbps.

Like in the conventional FCS or in the capacity based FCS, information on status

of the radio resources (muted or not) must be forwarded by the serving cell to all cells in

the active set. This information is carried in the FCS command message (see Table 9).

Note that the information on occupied resources is delivered to the cells in the active set

even in the conventional FCS. Therefore, the only difference in the overhead is in

delivery of the information on muting. The information on muting for each cell in active

set is represented by one bit (just on or off is indicated). Considering average size of the

Fast Cell Selection

76

active set around two cells (see Figure 45), the overhead due to our proposal is up to 0.2

kbps (2 cells reported once per 10 ms frame). Even if this message introduces additional

overhead, the overhead is increased only negligibly related to the conventional FCS

(difference of tens of bps). Contrary, the overhead is even slightly lower (again only

tens of bps) than in the case of the capacity based FCS since the active set size of our

proposal is lower (see Figure 45).

6.3 CONCLUSIONS

In this chapter, first, we investigate the performance of FCS and hard handover if

the small cells, connected to the network via either limited or unlimited backhaul, are

considered. The results show slight increase in the amount of mobility events (AS

updates) if FCS is used. On the other hand, FCS fully eliminates the interruption due to

the user mobility. Therefore, FCS is profitable for real-time services such as voice calls.

In term of the throughput, FCS introduces significant gain for all UEs for the unlimited

backhaul capacity, i.e., for the pico/microcells. This confirms observations presented by

other researchers for the scenarios with macrocells only as the only difference between

the macro and the pico/micro cells consists in the cell radius. On the other hand, the

throughput is improved by FCS only for the outdoor UEs offering low throughput up to

2 Mbps if the backhaul capacity is limited (that is, for femtocells). Therefore, if the

small cells are deployed, the conventional FCS can even decrease performance in term

of the throughput if the backhaul capacity is not considered in the active set

management. Hence, algorithms for mobility support should be aware of the available

capacity of the small cell backhaul to maximize the throughput of users.

Based on the observation of the FCS's performance, the algorithm related to FCS

active set management considering backhaul limitations introduced by deployment of

FAPs in networks is designed. The proposed algorithm is based on comparison of the

amount of MBS's radio resources consumed if a cell is included to the UE's active set or

not. The simulation results show notable increase in throughput for indoor as well as

outdoor users. Simultaneously, the amount of generated overhead is significantly

reduced by the proposal. Moreover, the proposed algorithm reduces the time, when

users are not fully satisfied with experienced capacity and delay.

As the simulations show, the most efficient active set always contains an MBS.

For indoor users, the closest FAP deployed in the same house should be included as

Fast Cell Selection

77

well. The amount of FAPs included in the active set together with MBS for outdoor

users depends on mutual distance between the UE and neighboring FAPs. Further, the

number of FAPs slightly varies with offered traffic level. In average, roughly 1.3 FAPs

and 1.5 FAPs are included in active set of outdoor UE for low/high traffic level and for

medium traffic (100kbps - 1500 kbps).

Temporary Access to Closed FAP

78

7 TEMPORARY ACCESS TO CLOSED FAP

This section addresses the problem of a dynamic management of a CSG list of a

closed FAP (denoted as CSG FAP). The goal is to ensure simple and easy management

of “adding” or “removing” new UEs, so called "Visiting UEs", to or from the CSG list.

The reference scenario is depicted in Figure 47. Users included on the CSG list of a

CSG FAP are denoted as CSG UEs.

Figure 47. Reference scenario for management of visiting users.

Each UE is aware of all CSG FAPs that this UE can access. These CSG FAPs are

included in each UE's CSG whitelist. The whitelist is a combination of Allowed CSG

list and Operator CSG list. The former one is under control of both the operator and the

user, while the latter one is under exclusive control of the operator (for more

information, see [38]). Both lists should be stored in the UE's USIM (Universal

Subscriber Identity Module). Each UE can access all CSG FAPs included on at least

one of the lists. Therefore, if the CSG FAP in the UE's range is listed in the whitelist,

the conventional procedures for connection control defined in [58] are performed. In

this paper, we focus on the scenario when the CSG FAP is not included in the UE's

whitelist but the UE still would like to access this FAP. In this case, the UE must obtain

permission from a subscriber of the CSG FAP. By the CSG FAP subscriber is

understood a user with the CSG FAP in its whitelist and with permission to allow/deny


79

access of the V-UEs. In the simplest way, the user who is the operator's signed

subscriber (denoted as Primary UE in this chapter) should be the user in charge of the

CSG list management. Besides the Primary UE, a list of potential users with permission

to control the CSG list of admitted UEs should be defined in case the Primary UE is out

of the CSG FAP range or if the Primary UE is willing to grant such rights to other

members of the CSG list (e.g., other members of the family or selected employees).

If a V-UE moves close to a FAP and the V-UE is able to receive and recognize an

identity of this cell, it also receives information about CSG status. This means, the V-

UE is able to determine whether this FAP utilizes closed or open access. This is

indicated by a CSG indication flag set to "true" broadcasted by each FAP together with

other information (see [36]). If the UE would like to perform handover to this cell, it

should conduct a measurement of the signal level received from this CSG FAP.

Handover can be performed even if no measurements are reported to the network.

However, this introduces a risk of the UE’s disconnection or QoS degradation if the

CSG FAP signal is interfering heavily to the UE connected to another cell. Therefore,

even if the measurement is optional, it is recommended to perform the measurement

before the handover initiation. The current 3GPP standards imply exclusion of CSG

FAP from signal measurement and reporting if no CSG FAP is in the UE’s whitelist

(see [10], [36]). Therefore, a modification enabling the UE to measure and to report

signal to the network even if no CSG FAP is included in its whitelist is necessary. For

the inclusion of such a CSG FAP in measurement and reporting, we introduce a new

flag MeasCSGFlag. The MeasCSGFlag is kept in the USIM of the UE along with the

whitelist. This flag can be set either manually by the V-UE, if it is willing to enter a

CSG FAP or automatically by the network if a strong interferer for a long time is noted

by the V-UE and the network expects handover to this FAP.

7.1 CONTROL PROCEDURE ENABLING ACCESS OF V-UES

In this section, the general framework of the management message flow is

outlined. Furthermore, two approaches, in-band and out-of-band, are described in more

detail.

7.1.1 GENERAL FRAMEWORK

The general overview of the proposed CSG management is depicted in Figure 48.

If a V-UE detects a CSG FAP, it can try to enter this FAP. In the conventional way, the


80

V-UE’s attempt to enter the CSG FAP without permission would be rejected as both the

FAP and the network consider this request as unjustified. Therefore, the request from

the V-UE must contain a new flag, "Not Allowed" ("NA"). This flag indicates that the

V-UE is aware of the fact that it cannot access this CSG FAP, and that the V-UE applies

for negotiation of access to the CSG FAP.

After that, the FAP, in cooperation with the network, finds an appropriate user

who has the right to accept or deny the V-UE request (i.e., the Primary UE or its

representative is found). The selection of the Primary UE is done only among CSG FAP

users with permission to grant the access. Among those UEs, the one with the highest

priority for CSG list management out of all CSG UEs in FAP's range is chosen. This

selected UE (shown as Primary UE in Figure 48), is asked if the V-UE can be admitted

to the CSG FAP. The Primary UE then either approves or rejects this request. If the

access of the V-UE is accepted, the Primary UE must provide an input for

authentication purposes and a Class of V-UE. The authentication input is understood as

a definition of access password for verification of the V-UE. The Class of V-UE stands

for set of limitations for the V-UE (e.g., bandwidth limit, overall amount of transferred

data, restriction of some applications or services, duration of the access, etc.). Note that

observance and enforcement of Class of the V-UE is in charge of the FAP.

The restrictions set by the Primary UE are then negotiated with the V-UE. Also,

the V-UE is asked to verify its identity by password. The handover to the CSG FAP can

be initiated only if the password entered by the V-UE is identical to the one provided by

the Primary UE and if the V-UE accepts all restrictions and conditions set in the Class

of V-UE.

Figure 48. General outline of the procedure for V-UE entering the CSG FAP.


81

This new scheme can introduce potential problem related to malicious attacks

when a UE could continuously try to enter a CSG FAP. This can be easily avoided by

definition of a minimum interval between two consecutive requests to enter the CSG

FAP issued by the same V-UE. Beside, also a blacklist of UEs with restricted access the

CSG FAP should be established. This blacklist should be under control of the CSG FAP

and the Primary UE.

Two options of management of the V-UE access are proposed in the following

subsections: In-Band (IB) and Out-Of-Band (OOB). The first one assumes signaling for

handling the V-UE entry within a conventional band used by the UE for all types of

communications (including data) with the network. The second one requires other radio

technology, such as Bluetooth, for the signaling.

7.1.2 IN-BAND APPROACH

The flow of control messages for admission of the V-UE to the CSG FAP with

utilization of IB is depicted in Figure 49. Both signaling over radio and backhaul are

illustrated. If the V-UE is able to detect the CSG FAP, it sends a request for access to

this FAP. The request is transmitted to the serving MBS since communication with the

FAP is not yet established. If the "NA" access is indicated by the V-UE, the MBS

forwards the request to the CSG FAP via backhaul. The FAP then transmits the V-UE

Request message to the Primary UE. This message contains only identification of the V-

UE to minimize redundant signaling overhead. Note that structure and detailed content

of all new required control messages is presented in the next section.

The Primary UE can either grant or deny the request using a V-UE Response

message. This message contains the ACK or NACK indication (grant or deny). If ACK

is present, then the Primary UE can define additional requirements or limiting

conditions for using the CSG FAP (Class of V-UE). Moreover, a password for

verification of the V-UE must be included in this message. This message is further

forwarded to the V-UE through the FAP, the FAP backhaul, and the serving MBS. For

security reasons, the password is not forwarded by the FAP. This means, the password

is delivered only from the Primary UE to the CSG FAP and then it is removed from the

message.

After reception of the V-UE Response, the V-UE enters either the password with

acknowledgment or rejection of the Class of V-UE set by the Primary UE. This


82

feedback is delivered to the CSG FAP via the serving MBS in V-UE Info message. Note

that the V-UE Response and V-UE Info messages can be exchanged more than once if

the agreement on Class of V-UE is not agreed in the first round. The CSG FAP then

compares both passwords. If both passwords are identical, the FAP confirms admission

of the V-UE to the Primary UE and to the network by means of a V-UE Confirm

message. Based on this message, the network includes the V-UE on the list of UEs with

access to this CSG FAP and handover can be initiated. To avoid a security risk, the

conventional authorization and security procedures are performed during handover. It

means, even if an attacker obtain permission from the Primary UE, it has to pass

network authentication and authorization procedure before it can communicate with the

network.

Figure 49. Flow of control messages for V-UE access using IB approach.

Since the temporary agreement on enabling the V-UE access to the CSG FAP

does not imply any commitments to allow access in the future, no update of the

whitelist in the V-UE is performed. After an expiration of the granted access, all new

records in the CLC must be deleted. Therefore, a timer must be run to ensure deletion of

such records. Update of the whitelist in USIM of the V-UE is necessary only if the

Primary UE indicates unlimited access grant for the V-UE (note that this is not

indicated in Figure 49 since we focus mainly on temporary access).


83

7.1.3 OUT-OF-BAND APPROACH

Another option for managing V-UE access is to use OOB communication since

nearly all mobile devices available at the market are equipped with a short-range

communication technology such as Bluetooth. If the V-UE comes to the vicinity of the

CSG FAP and if the Primary UE is in the range of OOB communication technology, the

V-UE can initiate the procedure via OOB by transmission of V-UE Request (see Figure

50). This message is sent via OOB directly to the Primary UE with the same content as

in the case of IB communication. The Primary UE can either accept or reject the V-UE

by a V-UE Response. In the case of accepting the V-UE request, a password and

additional limitations can be set by the Primary UE in the same way as for the IB

method. The confirmation of those requirements is sent by the V-UE together with the

password in V-UE Info. Again, the V-UE Response and V-UE Info messages can be

exchanged until an agreement on Class of V-UE is reached. If the OOB is used, the

Primary UE is responsible for verification of the V-UE authenticity. Like in Bluetooth

pairing, the password from the Primary UE to the V-UE is not transmitted via radio.

The Primary UE delivers the password to the visiting user personally (the primary user

says it or writes it down to the visiting user). Once the V-UE agrees to the conditions

defined by the Primary UE and both entered passwords match, the final

acknowledgment is sent to the V-UE in V-UE Confirm. At the same time, the Primary

UE informs the CSG FAP of the temporary inclusion of the V-UE to the list of UEs

with enabled access (V-UE Confirm) and of Class of V-UE (V-UE Response). The CSG

FAP forwards V-UE Confirm to the CLC via backhaul. Handover can be initiated by the

serving MBS after the reception of CLC update confirmation. Note that all conventional

authorization and security procedures are performed during handover to avoid security

risks as explained before for the IB approach.


84

Figure 50. Flow of control messages for V-UE access using OOB approach.

7.2 MANAGEMENT MESSAGES FOR VISITOR ACCESS

In this section, a content of new management messages and comparison of IB and

OOB are presented.

For both approaches of handling V-UE access, four new messages must be

designed: V-UE Request, V-UE Response, V-UE Info, and V-UE Confirm. Each message

starts with a message ID to distinguish its purpose. The second part of all messages is an

identification of the V-UE by 64-bits IMSI (International Mobile Subscriber Identity).

The V-UE Request message contains both IDs (message and V-UE). Optionally,

also a name of the V-UE assigned by the user can be included. This field is not

indicated in Table 15 as it just increase overhead and it is not necessary for successful

V-UE entry. The content of the V-UE Request message is presented in Table 15.

Table 15. Structure of V-UE Request message

Message field Size Description

Message ID TBD Identification of the message

ID of V-UE 64 bits Identification of the V-UE by IMSI

The second message, V-UE Response, is presented in Table 16. This message

contains identification of the message and identification of the V-UE like the V-UE

Request. Further, ACK/NACK of the access is indicated. If access is granted, a


85

password must be included for IB communication. For OOB, the password is told to

V-UE by Primary UE and it is carried neither in OOB not in 4G channels for data

communication. The length of the password field depends on the encrypting algorithm

and the password length. The password can be followed by optional conditions,

restrictions, or duration of the V-UE access defined in the Class of V-UE. The length of

this field depends on the amount of restrictions and conditions set by the Primary UE.

Table 16. Structure of V-UE Response message




ACK/NACK 1 bit ACK ... access of the V-UE enabled

NACK ... access of the V-UE disabled

Password Variable Password for verification of the V-UE

Class of V-UE Variable Defines restriction to the V-UE and duration

of granted access

The next message, V-UE Info, is presented in Table 17. This message is a

feedback from the V-UE to the V-UE Response. Beside the IDs of the message and the

V-UE, the additional field, ACK/NACK, is mandatory. It indicates whether the V-UE

accepts the condition for the FAP’s access defined by the Primary UE. If the conditions

are accepted, the field with the password is included just after the ACK/NACK field.

Table 17. Structure of V-UE Info message




ACK/NACK 1 bit ACK ... acceptation of Class of UE

NACK ... rejection of Class of UE

Password Variable Password for verification of V-UE's

The last message, V-UE Confirm, is presented in Table 18. This message ends the

process of granting the V-UE access to the CSG FAP. In addition to the message ID and

the ID of the V-UE, 9 bits with the FAP’s ID is included. We suppose to use the same

indicator as the Physical Cell ID (PCI). The PCI distinguishes up to 504 cells [58], thus

9 bits are required for the FAP identification. This ID is included just for CLC purposes

(CLC always contains pairs - ID of the CSG FAP and related ID of the admitted V-UE).


86

Finally, information on the duration of the access to the CSG FAP is presented in the

message. This information must be delivered to the network to ensure deletion of the

record from the CSG list in CLC after an expiration of the access grant.

Table 18. Structure of V-UE Confirm message




CSG FAP ID 9 bits Identification of the FAP, the same number

as the FAP's Physical Cell ID can be used.

Access grant

duration

TBD Information on duration of enabled access to

the CSG FAP.

If this field equals zero, unlimited access is

indicated.

Comparing IB and OOB, the latter one imposes a lower amount of signaling

overhead on radio interface and backhaul links than IB (five messages are transferred

via IB radio and backhaul instead of thirteen). Contrary, OOB requires enabled OOB

communication technology on both UEs (V-UE and Primary UE). Therefore, the OOB

could negatively influence the battery lifetime of both involved devices due to the need

of other additional radio communication technology. It means there is a trade-off

between battery life-time and signaling overhead. Nevertheless, the OOB is used only

for a very short time before entering CSG FAP (up to few minutes). As well, only

negligible overhead is generated by this procedure (up to few kilobits). Hence,

appropriate way can be arbitrary selected according to users and/or operators

preferences.

7.3 IMPACT OF THE TEMPORARY V-UE ACCESS ON THE

V-UE'S PERFORMANCE

The proposed management of the temporary V-UE access allows to change the

access mode for a V-UE. It means, a closed FAP becomes temporary the open FAP for

the V-UE. Therefore, change in performance of the V-UE corresponds to the difference

between the open and closed accesses. This problem has already been investigated, for

example, in [1] and results demonstrate a gain in throughput due to the open access if

the UE is close to the FAP. Therefore, we show only illustrative impact of the


87

temporary V-UE access on its SINR (Signal to Interference plus Noise Ratio) measured

by the V-UE.

For evaluation, we consider model with an MBS and a FAP deployed in mutual

distance denoted dMBS-FAP. The MBS and FAP transmit with 46 and 15 dBm

respectively. Signal from the MBS is propagated according to the Okumura-Hata model

while signal inside the building follows ITU-R P.1238 model as recommended by Small

Cell Forum for residential buildings in small to medium city [43]. The building is of a

rectangular shape with a size of 10 x 10 m. Carrier frequency of 2 GHz and noise with

density of -174 dBm/Hz are also considered for signal propagation. As the FAP is

placed in the middle of a building, wall attenuation of 10 dB is taken into account for

communication between the V-UE and the outdoor MBS.

Cumulative density function (cdf) of SINR experienced by the V-UE if the access

to the closed FAP is disabled or enabled is depicted in Figure 51. The SINR is derived

for 121 normally distributed positions of the V-UE inside the building and for 500

values of dMBS-FAP distance. The dMBS-FAP is also normally distributed in the range from 0

to 500 m. If the V-UE cannot access the closed FAP and stays connected to the MBS, it

suffers from heavy interference incurred by the FAP. In this case the observed SINR

varies only between -30 dBm and 12 dBm. Consequently, the V-UE is not able to

receive signal from the MBS with sufficient quality most of the time. However, if the

V-UE is temporarily admitted to the FAP, its SINR improves dramatically. To be more

specific, SINR measured by the V-UE is in a range of -10 dBm and 30 dBm.

Figure 51. SINR experienced by V-UE if temporary access is not enabled (dashed

blue line) and if the V-UE is enabled to access this FAP (solid red line).


88

As could be expected, higher dMBS-FAP distance reduces signal quality observed by

the V-UE from the MBS as shown in Figure 52. This figure depicts average SINR

measured by the V-UE at 121 normally distributed positions within the building.

According to the results presented in Figure 52, the temporary access of the V-UE

should be applied especially in the case when the signal from the MBS is weak

comparing to the signal of the FAP. On the other hand, if the FAP is close to MBS

(distance up to 60 m), it is better for the V-UE to stay connected to the MBS. This is in

compliance with fact that the deployment of the FAP is advantageous mainly for

location with weak signal quality from the MBS.

Figure 52. SINR experienced by V-UE over distance between MBS and FAP if

temporary access is not enabled (dashed blue line) and if the V-UE is enabled to access this FAP (solid red line).

7.4 CONCLUSIONS

This chapter introduces new procedure to enable temporary access of a visiting

user to the CSG FAPs. Contrary to the existing solutions, the new one is convenient for

frequent update and easy management of the CSG list. For this purpose, we have

defined chart flow of control messages as well as their content for IB and OOB way of

the V-UE access management. Signaling overhead introduced by the new procedure is

only few kilobits per access and can be neglected. Since both approaches still exploit

full conventional authorization and security procedures before the V-UE is admitted to

the CSG FAP, increased security risk is introduced by non of both ways of

management.

Conclusions and Future Work

89

8 CONCLUSIONS AND FUTURE WORK

This thesis is focused on management of problems related to the mobility of users

in mobile networks with small cells. Three major topics are addressed: hard handover,

fast cell selection, and temporary access to closed femtocells.

In the area of hard handovers, two algorithms for elimination of redundant

handovers are proposed. Both proposed schemes differ in its requirements on

modifications of current standards. While the first scheme, adaptive techniques,

exploits only conventionally observed and monitored parameters, the second one, ETG,

requires estimation of user's throughput. The second scheme significantly outperforms

the first one in both user's throughput and efficiency in elimination of redundant

handovers. However, it is at the cost of higher computational complexity and additional

signaling overhead. Nevertheless, both proposed schemes show higher performance

comparing to the conventional and competitive proposals. Performance of adaptive

techniques could be improved, in the future, by sensing capabilities of the femtocells. It

means, the maximum level of signal should be determined exactly according to the

measurement of the signal level directly by the femtocells. Future enhancement of ETG

can be achieved by advanced estimation of the signal evolution or interference.

Moreover, more precise estimation of time spend in the cell by each user based on

personnel characteristics and behavior of each user could further enhance performance

of ETG.

Further, the performance of FCS and hard handover if the small cells are deployed

in the networks is investigated. Analogically to the macrocells, FCS introduces

significant gain in throughput for all UEs if small cell backhaul is of unlimited capacity,

i.e., for the pico/microcells. However, the throughput is improved by FCS only for the

outdoor UEs offering low throughput if the femtocell backhaul capacity is limited.

Otherwise, the throughput is even degraded by FCS. Hence, the algorithm related to

Conclusions and Future Work

90

FCS active set management considering backhaul limitations introduced by deployment

of FAPs in networks is designed. The proposed algorithm is based on comparison of the

amount of MBS's radio resources consumed if a cell is included to the UE's active set or

not. The simulation results show significant gain in throughput for indoor as well as

outdoor users. Moreover, the proposed algorithm reduces signaling overhead related to

the active set management and the time when users are not fully satisfied with

experienced capacity and delay. The proposed algorithm can be further extended for

FAP's downlink power control to reduce interference from cells that cannot fulfill UE's

requirements.

Last part tackles the problem of temporary access of visiting UEs to the closed

FAPs. The proposed solution is convenient for frequent update and easy management of

the CSG list. To enable new management procedure, several new control messages are

proposed. Moreover, the chart flow of control messages for IB and OOB way of the V-

UE access management are designed as well. OOB approach requires enabled

additional communication technology (e.g. Bluetooth) on both involved devices.

However, it reduces signaling overhead in communication band.

Beside further incremental enhancement of individual algorithms and techniques

to improve their performance, the mobility management can adopt prediction

approaches considering a periodicity in users' behavior. It means, to exploit the fact that

users usually follows similar patterns in daily movement and daily traffic.

Also a problem of merging mobile communications with other technologies such

as cloud computing should be considered in future research. In this case, user's

movement could significantly influence computation or transmission of large amount of

data between the cloud and the user. In this case, management of routing of user's data

and data processing should be aware of the user's requirements on computational and

storage capacities of remote clouds. To that end, distribution of a centralized cloud

closer to the users (e.g. to small cells) can significantly improve user's experienced QoS.

Summary of Research Contributions

91

SUMMARY OF RESEARCH CONTRIBUTIONS

The habilitation thesis is focused on the support of user's mobility in networks

with small cells. The contributions of the thesis into the area of handovers are

following:

Chapter 5

� Proposal and evaluation of the adaptive techniques for minimizing negative

impact of handover on the user's throughput and improving efficiency in

elimination of redundant handovers.

- Related results are includes in following papers:

1. Z. Becvar - P. Mach, "Adaptive Hysteresis Margin for Handover in

Femtocell Networks," International Conference on Wireless and

Mobile Communications (ICWMC 2010), Valencia, Spain, 2010.

2. Z. Becvar - P. Mach, "Adaptive Techniques for Elimination of

Redundant Handovers in Femtocells," International Conference on

Networks (ICN 2011), St. Maarten, Netherlands, 2011.

3. Z. Becvar - P. Mach - M. Vondra, "Handover Procedure in

Femtocells," In Femtocell Communications and Technologies. IGI

Global, 2012, pp. 157-179, edited by R.A. Saeed, B.S. Chaudhari,

R.A. Mokhtar.

� Proposal and evaluation of the algorithm for handover decision exploiting

estimation of user's gain in throughput for minimizing negative impact of

handover on the user's throughput and improving efficiency in elimination of

redundant handovers


4. Z. Becvar - P. Mach, "On Enhancement of Handover Decision in

Femtocells," 4th IFIP Wireless Days, Niagara Falls, Canada, 2011.

5. Z. Becvar - P. Mach, "Estimation of Throughput Gain for

Handover Decision in Femtocells," submitted to journal Mobile

Information Systems in May 2012.

Summary of Research Contributions

92

Chapter 6

� Evaluation of the performance of the FCS in networks with small cells for

various types of users.


6. Z. Becvar - P. Mach, "On Enhancement of Handover Decision in

Femtocells," submitted to 5th IFIP Wireless Days, Dublin, Ireland,

2012.

� Proposal on active set management considering amount of MBS's resources

consumption if femtocells are deployed and included in active sets.


7. Z. Becvar - P. Roux - P. Mach, "Fast Cell Selection with Efficient

Active Set Management in OFDMA Networks with Femtocells,"

accepted for publication in EURASIP Journal on Wireless

Communications and Networking, 2012.

Chapter 7

� Design of the control procedure for temporary admission of visiting users to

closed femtocells to improve users signal quality.


8. Z. Becvar - P. Mach, " Management Procedure for Temporary

Access of Visiting Users to Closed Femtocells," submitted to

journal KSII Transactions on Internet and Information Systems in

September 2012.

As this thesis have been completed in frame of FP7 FREEDOM project, all results

are also included in deliverables D4.1 - "Advanced procedures for handover in

femtocells" and D4.2 - "Design and evaluation of effective procedures for MAC layer"

of the project. Both documents are available at www.ict-freedom.eu.

References

93

REFERENCES

[1]

G. de la Roche, A. Valcarce, D. Lopez-Perez, J. Zhang " Access control

mechanisms for femtocells," IEEE Communication Magazine, vol.48, no. 1, pp.

33-39, January 2010.

[2] Small Cells Forum, "Femtocells – Natural Solution of Offload," June 2010.

[3] D. Lopez-Perez, A. Valcarce, A. Ladanyi, G. de la Roche, and J. Zhang, "Intracell

Handover for Interference and Handover Mitigation in OFDMA Two-Tier

Macrocell-Femtocell Networks," EURASIP Journal on Wireless Communications

and Networking, 2010, 15 pages, 2010.

[4] P. Xia, V. Chandrasekhar, and J. G. Andrews, "Open vs. Closed Access

Femtocells in the Uplink," IEEE Transactions on Wireless Communications, Vol.

9, No. 12, 3798 - 3809, .2010.

[5] Small Cell Forum whitepaper, " Small cells – what’s the big idea? Femtocells are

expanding beyond the home," February 2012.

[6] V. Chandrasekhar, J. Andrews, A. Gatherer, "Femtocell Networks: A Survey,"

IEEE Communication Magazine, vol.46, no. 9, June 2008.

[7] S. Moghaddam, V. Tabataba, and A. Falahati, "New handoff initiation algorithm

(optimum combination of hysteresis and threshold based methods)," IEEE VTC

2000-Fall, September 2000.

[8] D. Rose, T. Jansen, T. Kurner, "Modeling of Femto Cells - Simulation of

Interference and Handovers in LTE Networks," IEEE VTC 2011-Spring, 2011.

[9] T. Jansen, I. Balan, J. Turk, I. Moerman, and T. Kurner, "Handover parameter

optimization in LTE self-organizing networks," IEEE VTC 2010-Fall, September

2010.

[10] 3GPP Technical Specification TS 36.133 v 10.6.0, "Technical Specification

Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-

UTRA); Requirements for support of radio resource management (Release 10),"

March 2012

[11] M. Zonoozi, P. Dassanayake, and M. Faulkner, "Optimum hysteresis level, signal

averaging time and handover delay," In IEEE VTC 1997, pp. 310 – 313, 1997.

[12] K.I. Itoh, S. Watanabe, J.S. Shih, and T. Sato, "Performance of handoff algorithm

based on distance and RSSI measurements," IEEE Transactions on Vehicular

Technology, Vol. 51, No. 6, 1460 – 1468, 2002.

[13] S. Lal and D. K. Panwar, "Coverage Analysis of Handoff Algorithm with

Adaptive Hysteresis Margin," In ICIT 2007, pp. 133 – 138, 2007.

[14] J.M. Moon and D.H. Cho, "Efficient Handoff Algorithm for Inbound Mobility in

Hierarchical Macro/Femto Cell Networks," IEEE Communications Letters, Vol.

13, No. 10, 755- 757, 2009.

[15] J.M. Moon and D.H. Cho, "Novel Handoff Decision Algorithm in Hierarchical

Macro/Femto-Cell Networks," In IEEE WCNC 2010, pp. 1- 6, 2010.

References

94

[16] H. Zhang, X. Wen, B. Wang, W. Zheng, and Y. Sun, "A Novel Handover

Mechanism between Femtocell and Macrocell for LTE based Networks," In

ICCSN2010, 2010.

[17] L. Barolli, "A speed-aware handover system for wireless cellular networks based

on fuzzy logic," Mobile Information Systems, Vol. 4, No. 1, pp. 1-12, 2008.

[18] G. Mino, L. Barolli, F. Xhafa, A. Durresi and A. Koyama, "Implementation and

performance evaluation of two fuzzy-based handover systems for wireless cellular

networks," Mobile Information Systems, Vol. 5, No. 4, pp. 339-361, 2009.

[19] H. Claussen, F. Pivit, and L.T.W.Ho, "Self-Optimization of Femtocell Coverage

to Minimize the Increase in Core Network Mobility Signalling," Bell Labs

Technical Journal, Vol. 14, No. 2, pp. 155–184, 2009.

[20] H.S. Jo, Ch. Mun, J. Moon, and J.G. Yook, "Self-Optimized Coverage

Coordination in Femtocell Networks," IEEE Transactions on Wireless

Communications, Vol. 9, No. 10, pp. 2977-2982, 2010.

[21] S. Y. Yun and D. H. Cho, "Traffic Density based Power Control Scheme for

Femto AP," In IEEE PIMRC 2010, pp. 1378-1383, 2010.

[22] Y. Choi and S. Choi, "Service Charge and Energy-Aware Vertical Handoff in

Integrated IEEE 802.16e/802.11 Networks," In IEEE INFOCOM 2007, 2007.

[23] P. Fülöp, S. Imre, S. Szabó and T. Szálka, "Accurate mobility modeling and

location prediction based on pattern analysis of handover series in mobile

networks," Mobile Information Systems, Vol. 5, No. 3, pp. 255-289, 2009.

[24] P. Bellavista, M. Cinque, D. Cotroneo, and L. Foschini, " Self-Adaptive Handoff

Management for Mobile Streaming Continuity," IEEE Transactions on Networks

and Service Management, Vol. 6, No. 2, 2009.

[25] J. Martinez-Bauset, J.M. Gimenez-Guzman and V. Pla, "Optimal Admission

Control in Multimedia Mobile Networks with Handover Prediction," IEEE

Wireless Communications, Vol. 15, No.5, 2008.

[26] 3GPP TR 25.433 v10.4.0, "Technical Specification Group Radio Access

Network; UTRAN Iub interface Node B Application Part (NBAP) signalling,

Release 10," September 2011.

[27] 3GPP TR 25.848 v4.0.0, "Technical Specification Group Radio Access Network;

Physical layer aspects of UTRA High Speed Downlink Packet Access, Technical

report," Release 4, March 2001.

[28] S.-W. Kim, Y.-H. Lee, "Adaptive MIMO Mode and Fast Cell Selection with

Interference Avoidance in Multi-cell Environments," In Fifth International

Conference on Wireless and Mobile Communications ICWMC '09, Cannes, La

Bocca, 23-29 August 2009.

[29] H.-H. Choi, J.B. Lim, H. Hwang, K. Jang, "Optimal Handover Decision

Algorithm for Throughput Enhancement in Cooperative Cellular Networks," In

IEEE VTC 2010-Fall, Ottawa, 6-9 September 2010.

[30] H.-H. Choi, "An Optimal Handover Decision for Throughput Enhancement,"

IEEE Commun. Lett. Vol. 14, No. 9, pp. 851-853, 2010.

References

95

[31] NTT DOCOMO, "Investigation on Coordinated Multipoint Transmission

Schemes in LTE-Advanced Downlink," In 3GPP TSG RAN WG1 Meeting

#55bis , R1-060298, Ljubljana, 12-16 January 2009.

[32] M. Feng, X. She, L. Chen, Y. Kishiyama, "Enhanced Dynamic Cell Selection

with Muting Scheme for DL CoMP in LTE-A," In IEEE VTC 2010-Spring,

Taipei, 16-19 May 2010.

[33] K.-W. Lee, J.-Y. Ko, Y.-H. Lee, "Fast Cell Site Selection with Interference

Avoidance in Packet Based OFDM Cellular Systems," In IEEE Globecom 2006,

San Francisco, 27 November - 1 December 2006.

[34] A. Das, B. Krishna, F. Khan, S. Ashwin, H.-J. Su, "Network controlled cell

selection for the high speed downlink packet access in UMTS," in IEEE WCNC

2004, 21-25 March 2004.

[35] H. Fu, D. Kim, "Scheduling performance in downlink WCDMA networks with

AMC and fast cell selection," IEEE Trans. Wireless Commun., Vol. 7, No. 7, pp.

2580-2591, 2008.

[36] 3GPP TS 36.304 v10.5.0, "3rd Generation Partnership Project; Technical

Specification Group Radio Access Network; Evolved Universal Terrestrial Radio

Access (E-UTRA); User Equipment (UE) procedures in idle mode (Release 10),"

Mar. 2012.

[37] 3GPP TS 36.133, v10.6.0, "3rd Generation Partnership Project; Technical


Access (E-UTRA); Requirements for support of radio resource management

(Release 10)," Mar. 2012.

[38] G. Horn, "3GPP Femtocells: Architecture and Protocols," Qualcomm, Sept. 2010.

[39] ITU-R Document 5D/TEMP/89r1, "Draft new Report ITU-R M.[IMT.TECH],

Requirements related to technical system performance for IMT-Advanced radio

interface(s), " 2008.

[40] R. Srinivasan, et. al, "IEEE 802.16m Evaluation Methodology Document

(EMD)," Description of Evaluation of IEEE 802.16m System Performance, Rev.

IEEE 802.16m-08/004r2, July 2008.

[41] T. Camp, J. Boleng, V. Davies, "A Survey of Mobility Models for Ad Hoc

Network Research," Wireless Communications & Mobile Computing, Vol. 2, No.

5, 2002.

[42] ETSI UMTS 30.03, "Selection Procedures for the Choice of Radio Transmission

Technologies of the UMTS," ETSI Technical Report, 1998.

[43] Small Cell Forum, "Interference management in OFDMA femtocells," March

2010.

[44] G. Vivier, A. Agustin, J. Vidal, O. Muñoz, S.Barbarossa, L. Pescosolido, et al.,

"Scenario, requirements and first business model analysis," Deliverable D2.1 of

ICT-248891 STP FREEDOM project, June 2010.

[45] C. Yu, W. Xiangming, L. Xinqi, and Z. Wei1, "Research on the modulation and

coding scheme in LTE TDD wireless network," In ICIMA 2009, pp. 468 - 471,

2009.

References

96

[46] 3GPP TS 36.211 v 9.0.0, “3rd Generation Partnership Project; Technical


Access (E-UTRA); Physical Channels and Modulation,” Dec. 2009.

[47] Z. Becvar, J. Zelenka, “Implementation of Handover Delay Timer into WiMAX,”

Proc. of 6th Conference on Telecommunications (ConfTele 2007), Peniche,

Portugal, May 2007.

[48] N. P. Singh and B. Singh, “Performance of Soft Handover Algorithm in Varied

Propagation Environments,” World Academy of Science, Engineering and

Technology, vol. 45, 2008.

[49] ITU-R P.1238-6 Recommendation, “Propagation data and prediction methods for

the planning of indoor radiocommunication systems and radio local area networks

in the frequency range 900 MHz to 100 GHz,” 2009.

[50] Z. Becvar, P. Mach, "On Enhancement of Handover Decision in Femtocells," In

IFIP Wireless Days 2011, Niagara Falls, Canada, 2011.

[51] K. Kobayashi and T. Katayama "Analysis and Evaluation of Packet Delay

Variance in the Internet," IEICE Transactions on Communications, Vol. E-85B,

No. 1, pp. 35 - 42 , 2002.

[52] H. Hariyanto, R. Wulansari, T. A. Nugraha, H. Ahmadi, A.K. Widiawan, J.

Stéphan, Y. Corre, A. Cordonnier, R. Charbonnier, "Trial report," Deliverable

D6.2.1 of ICT-248891 STP FREEDOM project, February 2012.

[53] W. Daamen and S.P. Hoogendoorn, "Free Speed Distributions for Pedestrian

Traffic," presented at Transportation Research Board, 85th Annual Meeting, 2006.

[54] S. Sesia, I. Toufik, and M. Baker, LTE–the UMTS long term evolution: From

Theory to Practice. West Sussex: Wiley, 2009.

[55] H. Claussen, L.T.W. Ho, and L.G. Samuel, "Self-optimization of Coverage for

Femtocell Deployment," IEEE Wireless Telecommunication Symposium WTS

2008, pp. 278-285, April 2008.

[56] ITU-R M.2135 Recommendation, "Guidelines for evaluation of radio interface

technologies for IMT-Advanced," 2008.

[57] 3GPP TS36.331, v10.5.0, "3rd Generation Partnership Project; Technical


Access (E-UTRA); Radio Resource Control (RRC); Protocol specification

(Release 10)," Mar. 2012.

[58] 3GPP TS 36.300, v10.7.0, "3rd Generation Partnership Project; Technical


Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-

UTRAN); Overall description; Stage 2 (Release 10)," Mar. 2012.

Appendix

97

APPENDIX

VARIATION OF TIME IN FEMTOCELL

The determination of limits for error in estimation of tc is as follows. The distance

covered by j-th user in the femtocell is j,davg,fj,f dd ∆±= ; where avg,fd corresponds to

average distance covered by all users in the FAP's area and j,d∆ represents distance

deviation of j-th user's. Further, the speed of j-th user is j,vavg,jj vv ∆±= ; where avg,jv is

the average speed of pedestrians and j,v∆ stands for the speed variation. Since only

pedestrians are considered and since the mean speed of users is normally distributed

along 1ms34.1 − with 1max,j,v ms37.0 −±=∆ , i.e., 1

j ms37.034.1v −±= according to [53].

In compliance with above mentioned, average tc is defined as:

avg,javg,favg,j,c v/dt = . In relation to environment in femtocell, the lower and upper limit

for tc can be defined. The simplest case of infrastructure deployment is represented by a

single direct street as depicted in Figure 53.

Figure 53. Notation for determination of tc limits.

The real tc of individual user moving along direct street as depicted in Figure 53 is

limited from lower boundary to:

( ) ( ) ( ) 71.1/dv/dt j,davg,fmax,j,vavg,jj,davg,fmin,c ∆∆∆ −=+−= (40)

The tc,min depends on the position of a street in relation to the FAP radius. The

street is of a width w∆ and its borders are in distances D2 and D1 from the cell edge.

Appendix

98

Assuming the direct movement of users along the street, then tc,min is related to D2. The

distance df,2 covered by a user in the femtocell along the path distanced D2 from the cell

edge, is equal to ( )2

2f2f2,f Drr2d −−= . Therefore the tc,min as a function of D2 is:

( ) ( ) 71.1/Drr2v/dt2

2f

2

fmax,j,vavg,j2,fmin,c −−=+= ∆ (41)

The upper bound for tc is derived analogically to (40) assuming

97.0/r2d f1,f = (see Figure 53):

( ) ( ) ( ) 97.0/r297.0/dv/dt fj,davg,fmax,j,vavg,jj,davg,fmax,c =+=−+= ∆∆∆ (42)

Dependence of tc,min and tc,max over D2 is shown in Figure 54. This figure is

depicted for condition f1 rD = , which corresponds to the maximum df,1 ( f1,f r2d = ) and

thus to the worst case scenario. As Figure 54 shows, the variation of tc is up to roughly

2.1 multiple of the cell radius. This maximum variation occurs if the area of user’s

movement (a street or a sidewalk) covers at least a half of the cell radius ( fw r=∆ ).

However, the variation of tc is still significantly lower than in the case of the MBS since

ffemtomax,cB

mcromax,c r1.2t;r1.2t ×=×= and rB>>rf. Thus we can declare femto

max,cmacro

max,c tt >> .

Figure 54. Deviation of tc,min and tc,max over relative distance of users’ path from the

FAP’s position.

In more complex situation, the users are not limited to the direct movement. Their

movement is influenced by other factors such as a deployment of streets in the cell,

position of points of interests, users’ behavior, etc. All these factors can be represented

by function ξ . Further, the time in cell is affected by a probability TC that the user will

Appendix

99

stay longer in the cell, e.g., due to turn away from a direct movement or due to stop.

Therefore, TC is related to the ξ . Neither ξ nor TC can be easily determined. However,

both are clearly proportional to the cell radius as larger cell can cover more complex

infrastructure lay-out (e.g., more street crosses, more points of interests, etc.).

Therefore, the probability TC is significantly higher for larger cells:

)r(f);(fTC)r(f);(fTC frBr fB==>>== ξξξξ (43)

Above mentioned shows that the dispersion of minimum and maximum time in

cell is significantly lower for cells with low radius.

czech technical university in prague faculty of …etg estimation of throughput gain fap femto...

Documents